docs(notices): correct the Jello dates and put shipped upgrades in the past tense

[eddiemessiah]

The hole, and the fix

Several archived upgrade notices had a wrong date or still read as if the upgrade had not happened yet. AGENTS.md asks notices to state what is true now. This PR makes date and tense fixes, and turns "latest release" lists into time-bound release lists that point to Network config for current versions. It does not rewrite any notice.

Dates

• Jello notice (operate/notices/archive/jello-upgrade.mdx): the Celo Sepolia activation date was "Wed, Oct 5, 2025". The linked L1 transaction was mined on 2025-11-05 11:16:00 UTC, and the Jello spec page already says Nov 05. Changed to "Wed, Nov 5, 2025". (Oct 5, 2025 was also a Sunday.)
• Jello spec (operate/specification/upgrades/jello.mdx): the mainnet row said 10:24:22 UTC. The upgrade transaction in that row was mined in a block with timestamp 10:24:11 UTC. Changed to 10:24:11.
• L1 Fusaka notice: replaced "Expected early December 2025 … We will update this notice" with the actual activation time, Wednesday, December 3rd, 2025 at 21:49:11 UTC, in one sentence with the Sepolia date.

Tense and activation banners

• Jovian notice: "This page will be kept updated…" is now "Jovian has been activated.", which matches the Jello notice. Fixed the "Actvation" typo. The withdrawal Note, withdrawal steps 1-3, the operator upgrade line, the genesis/rollup re-download line, the "logged their configurations" check and the log index rebuild note are past tense. Step 4 (resubmit an invalidated proof) is kept as an instruction because it still applies.
• Isthmus notice: "This page will be kept updated…" is now "Isthmus has been activated." The operator upgrade line, the EigenDA proxy v1.8.2 line and the "logged their configurations" check are past tense. Removed the <Warning> that sent readers with a stuck Alfajores node to the Operators FAQ. That FAQ entry ("My Alfajores node stalled at the Isthmus hardfork block (49908280)") was deleted in docs: delete _deprecated/, prune dead-end redirects, add orphan-page check to CI #2279, so /operate/operators/faq no longer answers the question the warning asked.
• EigenDA v2 notice: removed "This page will be kept updated…". "is activated on the sequencer" is now "was activated on the sequencer". The operator upgrade line is past tense.
• Ice Cream spec (operate/specification/upgrades/ice-cream.mdx): "the Celo sequencer will use EigenDA v2 … must upgrade it's EigenDA proxy" is now "the Celo sequencer uses EigenDA v2 … nodes following the network had to upgrade their EigenDA proxy".
• L1 Fusaka notice: the L1-node instructions and the Note are past tense ("The Fusaka upgrade required no changes for Celo users", sourced from the page's own L2 client section, which says no Celo client upgrade was required).
• Celo Sepolia launch notice: removed "This page will be kept updated…". The intro, phase list, node-provider lines and the Longevity table cell no longer announce the Alfajores and Baklava retirement as future. They give 30 September 2025 as the end of support (from the forum post) and leave out a Holesky shutdown date, since the EF tied that to two weeks after Fusaka finalized on Holešky. The finished transition period carries the same ✅ as the phases above it.

Release lists

• Jello, L1 Fusaka, Celo Sepolia launch: "We recommend the latest released versions" is now "The versions released for this upgrade were …" (on Jello: "The Celo client versions recommended at the time were …", since those tags predate Jello) with the versions each page already listed, plus "For current versions see Network config."
• Jovian, Isthmus: the heading "Update to the latest release" is now "Releases for Jovian" / "Releases for Isthmus". The lead-in is past tense ("The release contained the activation timestamps …") and ends with the same Network config link. The version numbers are unchanged. No page in the repo linked to the old #update-to-the-latest-release anchor.

All eight files keep CRLF line endings. The diff against main contains only the changed lines (39 insertions, 58 deletions).

What this does NOT do / residual risk

• The Alfajores and Baklava rows in the Ice Cream and Isthmus activation lists stay. They record when the upgrade ran on those networks.
• "Users should be aware of the following impacts" (Jovian) and "This update requires some changes to the configuration" (Jovian EigenDA proxy, and the matching EigenDA v2 line) are left as they are. They describe the change itself, not a future event.
• The Ice Cream activation dates were not re-verified on chain. The Baklava and Alfajores chains are retired, and the notice says the upgrade was activated on the sequencer.

Judgement calls

• Removing the Isthmus FAQ warning instead of rewording it. The alternative was to inline the old fix (rewind op-geth with debug_setHead 49908270). That fix applies only to Alfajores, and AGENTS.md says not to reference Alfajores. This is a 4-line deletion and easy to revert if the maintainer wants the warning back with inlined text.
• Keeping the version lists on Jovian and Isthmus. They are the versions that carried the activation timestamps, so they stay as a record under a heading that names the upgrade. Current versions live on Network config.

Issues

Refs: none filed. These problems were found during a docs audit.

Stacking / conflicts

Branched off main (50ed71d). git merge-tree against current main is clean. #2346 (Alfajores references) touches none of these eight files.

Merge order: merge after #2376 (or together with it). This PR now points readers to Network config for current versions, and #2376 brings the versions on that page in line with the compose repo.

Verification evidence

1. Jello Celo Sepolia activation, L1 tx 0x736deb75…b31a on Ethereum Sepolia (https://ethereum-sepolia.publicnode.com, plus the Blockscout API)

tx 0x736deb757f4708eeafecc961f36e14f3711c1d9f944e45c933b666329743b31a blockNumber=0x91f586 (9565574)
block timestamp=0x690b31f0 (1762341360) => Wed 2025-11-05 11:16:00 UTC

GET https://eth-sepolia.blockscout.com/api/v2/transactions/0x736deb…b31a
"status":"ok" "method":"setRespectedGameType" "timestamp":"2025-11-05T11:16:00.000000Z" "block_number":9565574  (result: "success")

Jello is an L1 contract change (setRespectedGameType on the L1 OptimismPortal proxy), so its time comes from the L1 block, not from a Celo Sepolia L2 block.

2. Jello mainnet, L1 tx 0x5fb3f222…c44b on Ethereum mainnet (https://ethereum-rpc.publicnode.com, plus Blockscout)

tx 0x5fb3f2225dd2ba91efe941d4c9151df120ab06f6611b7b494bdecc96ff84c44b blockNumber=0x16dee7c (23981692)
block timestamp=0x69394a4b (1765362251) => Wed 2025-12-10 10:24:11 UTC
receipt: "blockNumber":"0x16dee7c" "status":"0x1"

GET https://eth.blockscout.com/api/v2/transactions/0x5fb3f222…c44b
"status":"ok" "method":"setRespectedGameType" "timestamp":"2025-12-10T10:24:11.000000Z" "block_number":23981692

3. Fusaka on Ethereum mainnet. Source: https://blog.ethereum.org/2025/11/06/fusaka-mainnet-announcement:

the announcement gives epoch 411392 / slot 13,164,544, "December 3, 2025 at 21:49:11 UTC". The Sepolia date matches https://blog.ethereum.org/2025/09/26/fusaka-testnet-announcement: **Sepolia** | 8,724,480 | 2025-10-14 07:36:00 | 1760427360.

On-chain check that mainnet is past Fusaka (eth_config, EIP-7910, on https://ethereum-rpc.publicnode.com):

current {"activationTime":1767747671,"utc":"2026-01-07T01:01:11.000Z","blobSchedule":{"target":14,"max":21,"baseFeeUpdateFraction":11684671},"P256VERIFY":"0x0000000000000000000000000000000000000100"}

The P256VERIFY precompile at 0x…0100 is Osaka-only. The current fork is BPO2, which is scheduled after Fusaka.

4. Jovian activation timestamps (unchanged, confirmed on chain). These timestamps match operate/specification/upgrades/jovian.mdx. The block header extraData switches from the version-0 format to the version-1 format (with the min base fee) exactly at the stated timestamp:

Celo mainnet (https://forno.celo.org)
63058029 1774958787 2026-03-31T12:06:27.000Z extraData=0x000000019000000005
63058030 1774958788 2026-03-31T12:06:28.000Z extraData=0x0100000190000000050000000000000000

Celo Sepolia (https://rpc.ankr.com/celo_sepolia, same via https://celo-sepolia.drpc.org; forno serves only the last ~10k blocks)
"number":"0x1384598" "timestamp":"0x69b9432c" "extraData":"0x000000019000000005"
"number":"0x1384599" "timestamp":"0x69b9432d" "extraData":"0x0100000190000000050000000000000000"
(0x69b9432d = 1773749037 = Tue 2026-03-17 12:03:57 UTC)

5. Isthmus FAQ target is gone

$ git show 762173a | grep -B1 -A1 49908280     # #2280, Legacy tab removal
-If you're encountering a stuck node after Alfajores hardfork block (49908280), see the [FAQ](/legacy/faq#my-alfajores-node-stalled-at-the-isthmus-hardfork-block-49908280).
+If you're encountering a stuck node after Alfajores hardfork block (49908280), see the [FAQ](/infra-partners/operators/faq).

$ git show 3604c62 | grep -A2 'My Alfajores node stalled'   # #2279 deleted the entry
-### My Alfajores node stalled at the Isthmus hardfork block (49908280)
-
-If you reached the hardfork block before upgrading to v2.1.0 your node can get stuck.

$ grep -rni "stalled at\|49908280" --include=*.mdx operate/operators/
(no output; operate/operators/faq.mdx has a single "## Mainnet" section with no Isthmus or Alfajores entry)

6. No stale strings left

$ grep -rn "Oct 5, 2025\|10:24:22\|Actvation\|Expected early December" --include=*.mdx .
(no output)
$ grep -rln "will be kept updated" --include=*.mdx .
(no output)
$ grep -rn "update-to-the-latest-release" .   # old Jovian/Isthmus heading anchor
(no output)
$ grep -nE " will |[Mm]ust |[Nn]eed to " <the 8 files in this PR>
operate/notices/archive/jello-upgrade.mdx:29: Node operators did **not** need to upgrade ...   (past tense already)

7. Local preview (npx mint dev, head 1f61865; the follow-up f7b2da5 changes wording only on the Celo Sepolia launch, Jello and EigenDA v2 pages, re-served locally with the new text and 0 console errors)

Previewed the three pages changed in the last commit (Jovian notice, Isthmus notice, Ice Cream spec) plus Network config, at 1280px and 412px wide.

• All four pages return 200. The new headings (#releases-for-jovian, #releases-for-isthmus) render, and the Network config link resolves.
• The new text is present in the served HTML, and none of the old phrases ("Update to the latest release", "will log their", "must re-download", "will use EigenDA", "it's EigenDA") remain.
• No horizontal overflow at either width (page scroll width equals the viewport at 412px). The Jovian heading fits on one line at 412px.
• Console errors: 0 on each page.

8. npx mint broken-links (on Windows)

Compared a run on this branch against a run on current origin/main. For the files in this PR the only differences are:

• removed: operate/notices/archive/isthmus-upgrade.mdx → /operate/operators/faq (the deleted warning)
• added: /operate/operators/network-config from the five notices that now link to it

On Windows the CLI reports almost every internal link as broken (backslash path handling). The same run lists /operate/operators/run-node as broken 22 times. operate/operators/network-config.mdx exists on main, is in docs.json, and serves 200 in the preview, so the added entries are the same false positive. CI on Linux is the authoritative run.

Remaining ops steps

• Merge after or with docs(operators): sync image versions, Grafana password and snapshot default with the compose repo #2376 (see Stacking).

Checklist

• Title is the commit message I want on main
• Ships the test that fails on pre-fix code; fixtures are real captured data; fakes honour their arguments (n/a, docs only)
• Covers the seam, not only pure functions; any stated guarantee tested on its failure path (n/a)
• Error responses audited for what they leak; reads on write paths bounded at the query (n/a)
• Judgement calls / bundled product changes flagged above (or "none")
• lint / typecheck / test pass locally on the current head (n/a; local preview and mint broken-links above)
• Re-read acceptance criteria of every Closes issue (no Closes)
• Development sidebar links match Closes/Refs above (none)
• README / runbook / .env.example / examples / error strings updated for the world this creates (n/a)
• Lockfile touched (no)
• Wallet/provider tree touched (no)
• UI touched (no)
• Money/security path (no)
• No secrets in the diff
• Questions for the maintainer marked clearly at the end (or "none")

Questions for the maintainer: keep the Isthmus FAQ warning deleted, or bring it back with the debug_setHead fix inlined?

🤖 Generated with Claude Code

[GigaHierz]

head 72118e0 · API headRefOid == local refs/pull ref (re-fetched); 1 commit, base 50ed71d, 1 behind main (a12466b, the home/ to learn/ rename); merge-tree against origin/main is clean; mint broken-links --check-redirects and check-orphans.sh pass on the merge; CI green (Documentation Validation success); rendered all 7 changed pages at 1280 and 412, 0 console errors, no horizontal overflow

Verdict: REQUEST-CHANGES

Thank you for this one. I rebuilt every date from primary sources and all the numbers you changed are right, including the two that were off by a month and by eleven seconds. The remaining findings are about tense: a few lines the PR now sits directly above still read as future, which is the thing this PR set out to fix.

Findings

1. Medium · operate/notices/archive/isthmus-upgrade.mdx:42, jovian-upgrade.mdx:48, eigenda-v2-upgrade.mdx:31 · "will need to upgrade ... before the activation dates" now contradicts the line the PR just changed

Each page now opens with "has been activated" or "was activated", and a few paragraphs later still tells operators to upgrade before the activation date ("Node operators will need to upgrade to the respective Isthmus releases before the activation dates", the Jovian equivalent, and "need to upgrade the EigenDA proxy ... before the activation date"). The PR text says these were left deliberately because they describe what the upgrade required.

Failure scenario: an operator or an AI assistant reads the Jovian page today, sees both "has been activated" and "will need to upgrade ... before the activation dates", and cannot tell whether an action is still pending on Celo Sepolia or mainnet (both have activated).

Fix shape: past tense with the same meaning, for example "Node operators needed to upgrade to the Jovian releases before the activation dates." One-word edits on three lines.

2. Medium · operate/notices/archive/celo-sepolia-launch.mdx:6, :20, :22, :35, :71 · the page still announces September 2025 events as future

The PR removed the "you are here" marker and the "will be kept updated" line, but the intro still says Celo Sepolia "will replace Alfajores when Holesky sunsets in September 2025" and Baklava "will also sunset", the phase list says "Sep 30, 2025: Planned Alfajores and Baklava sunset", and the call to action says "Please support both Alfajores and Celo Sepolia" (also line 35 and the "Sunset planned Sept 2025" table cell). Today is 2026-10-06. Alfajores is retired per AGENTS.md section 5.

Failure scenario: a node provider reads "Please support both Alfajores and Celo Sepolia during the transition period" and tries to run an Alfajores node that no longer has a network.

Fix shape: either rewrite those five lines in past tense in this PR (they are on the page the PR already edits), or, if the intent is to leave Alfajores wording to #2346, file a follow-up issue naming these lines and link it from the PR body. The review rule is that a deferred item needs a filed issue, not a prose promise.

3. Low · operate/notices/archive/celo-sepolia-launch.mdx:19 · the transition-period line is now the only finished phase without the check mark

The earlier four phases carry the check mark; "Aug 14, 2025 - Sep 14, 2025: Transition period" now has none, so a reader can take it as not finished. You flagged this as a judgement call.

Fix shape: add the check mark (one character), matching the lines above it.

4. Low · operate/notices/archive/jello-upgrade.mdx (For Node Operators), l1-fusaka-upgrade.mdx (L2 Celo client), celo-sepolia-launch.mdx (Release versions) · "We recommend the latest released versions: op-node v2.1.0 ..." is present tense and no longer true

These pages are in the set this PR puts in the past tense, but each still recommends "the latest released versions" with op-node v2.1.0, op-geth v2.1.2 and eigenda-proxy v1.8.2. The latest op-node release is celo-v2.2.3 (2026-10-05), compose now runs eigenda-proxy v2.6.0, and op-geth support has ended (see the op-geth notice). This is also at odds with the "document what is true now" rule, and with the #2376 sync of the operator pages.

Failure scenario: an operator reading the Jello page installs op-node v2.1.0 because the page says it is the latest.

Fix shape: past tense with no "latest": "The versions released for this upgrade were op-geth v2.1.2, op-node v2.1.0 and eigenda-proxy v1.8.2. For current versions see Network config." (the link target exists). Same one-line pattern on the three pages.

5. Low · operate/notices/archive/l1-fusaka-upgrade.mdx:11-12 and :18 · date now past, but the surrounding present and future wording remains

The mainnet line now has the exact activation time, but the line ends "affects Celo Mainnet" in present tense, the Note says "no changes are expected", and the next section still reads as pre-fork instructions ("You need to make sure that your L1 nodes are updated"). The PR declines to restate the "no changes" claim for lack of a source; the page's own L2 section does state "There are no upgrades required for the Celo L2 client", which is the source.

Fix shape: "Fusaka activated on Ethereum Mainnet on Wednesday, December 3rd, 2025 at 21:49:11 UTC and on Ethereum Sepolia on Tuesday, October 14th, 2025 at 07:36:00 UTC; it affected Celo Mainnet and Celo Sepolia respectively", and Note: "The Fusaka upgrade required no changes for Celo users." Only if you are comfortable with it, since the evidence for it is the page's own L2 client section plus Celo mainnet running normally after the fork; otherwise leave the Note and file an issue.

Verified good

All timestamps converted to UTC by me, from sources independent of the PR text:

• Jello Celo Sepolia: tx 0x736deb75...b31a is in Ethereum Sepolia block 9,565,574, header timestamp 1762341360 = Wed 2025-11-05 11:16:00 UTC (my own RPC read plus Blockscout 2025-11-05T11:16:00Z). It is a setRespectedGameType call to 0x44ae3d41... which equals OptimismPortalProxyAddress in the Celo Sepolia deployment-l1.json. The Celo forum post "Introducing the Jello Hardfork: OP Succinct Lite Now Live on Celo Sepolia" is dated 2025-11-05. Old text "Oct 5, 2025" was wrong (and a Sunday); "Wed, Nov 5, 2025" is right and now agrees with the spec page.
• Jello mainnet: tx 0x5fb3f222...c44b in Ethereum block 23,981,692, Blockscout timestamp 2025-12-10T10:24:11Z, a setRespectedGameType call to 0xc5c5D157... which equals OptimismPortalProxy in the mainnet deployment-l1.json. The forum thread "Jello will activate on Celo Mainnet on December 10, 2025" and the activation post of 2025-12-10 agree on the day. 10:24:11 is correct; the old 10:24:22 does not match the block. (My own mainnet RPC call failed on DNS, so the mainnet check used Blockscout, not a direct node read.)
• Fusaka: Ethereum mainnet slot 13,164,544 = epoch 411,392 x 32; genesis 1606824023 + 12 x 13164544 = Wed 2025-12-03 21:49:11 UTC; the Ethereum Foundation announcement states "December 3, 2025, 21:49:11 UTC" (fetched). Sepolia line: 1760427360 = Tue 2025-10-14 07:36:00 UTC, matches slot 8,724,480. Weekdays in the page are correct.
• Jovian (unchanged, cross-checked): rollup.json jovian_time is 1774958788 = Tue 2026-03-31 12:06:28 UTC on mainnet and 1773749037 = Tue 2026-03-17 12:03:57 UTC on Celo Sepolia; both match the notice and specification/upgrades/jovian.mdx.
• Isthmus (unchanged): mainnet isthmus_time 1752073200 = Wed 2025-07-09 15:00:00 UTC, matches the notice. The Alfajores and Baklava chains are retired and could not be checked.
• Ice Cream dates (Jul 30, Aug 20, Sep 10, 2025 15:00) were not re-verified; they match specification/upgrades/ice-cream.mdx, and the PR does not change them.
• Isthmus FAQ warning removal: after the PR no page on current main carries the Alfajores stalled-node entry; removing a link to a FAQ section that no longer exists is right.
• No notice now contradicts a specification page: Jello notice and jello.mdx agree on both dates and now on 10:24:11; Jovian, Isthmus and Ice Cream notices agree with their spec pages.
• "Tense of shipped upgrades": nothing not yet shipped was moved to the past. All seven upgrades are verifiably active (rollup config times are in the past; Fusaka is past; Jello is past).
• Tooling: mint broken-links --check-redirects and scripts/check-orphans.sh pass on a merge into current main; the repo-wide grep for the old strings ("Oct 5, 2025", "10:24:22", "Actvation", "Expected early December") returns nothing; "will be kept updated" no longer appears. Not checked: a Celo-side config.json for Celo Sepolia returned a non-JSON response, so I used rollup.json for both networks; Jello is an L1-only change, so neither file carries a Jello time.

What the PR got right

• Using the L1 block time for Jello, with the reasoning stated (it is an L1 contract change), is the right source and the evidence reproduces exactly.
• Catching the Sunday date, the 11-second drift, the "Actvation" typo and the dead FAQ link in one pass.
• A tight diff (7 files, 8 insertions, 14 deletions, CRLF preserved) and an honest "does NOT do" list. The Alfajores carve-out to avoid colliding with another PR is a sensible instinct.
• Not restating the Fusaka "no changes" claim without a source is exactly the right caution.

Merge-order hazards

• #2376 and #2345 do not touch these files; git merge-tree of #2345 against this PR is clean. #2376 merges in any order.
• #2346 (Alfajores references), per the PR text, touches none of these seven files; if it later edits the Celo Sepolia launch notice, it will conflict with findings 2 and 3 once fixed. Whoever merges second should expect that.
• #2367 (home/ to learn/) is already on main and this PR is only under operate/, so no rebase conflict. #2368 (build/ move) has not landed and does not touch these files.
• Findings 4 and the version numbers there depend on the #2376 sync; do not copy version numbers into the notices from either PR, link to Network config instead.

[palango]

I checked every date this PR changes, and they're all right. Jello mainnet is Ethereum block 23981692, timestamp 1765362251, which is 2025-12-10 10:24:11 UTC, and the transaction goes to the OptimismPortalProxy that superchain-registry lists for Celo. For Jello on Celo Sepolia, Blockscout shows Ethereum Sepolia block 9565574 at 2025-11-05 11:16:00 UTC. Fusaka mainnet is slot 13164544, and 1606824023 + 12 × 13164544 = 1764798551, which is 2025-12-03 21:49:11 UTC, matching the EF announcement. The Jovian and Isthmus times match superchain-registry. Dropping the Isthmus FAQ warning is right, since the FAQ no longer has that entry.

I agree with GigaHierz's points. The main one is that the pages now say the upgrade has happened while still telling operators to upgrade before the activation date (isthmus line 42, jovian line 48, eigenda-v2 line 31). If you rewrite the Jovian withdrawal section in the past tense too, please keep step 4 about resubmitting an invalidated withdrawal proof, because that still applies to anyone who never re-proved. On the Alfajores lines in the Celo Sepolia notice: my #2346 doesn't touch this file, so the two won't collide. I'm fine with the history staying, but "Please support both Alfajores and Celo Sepolia" is an instruction nobody can follow now, and I'd rather fix it here than leave it.

One small thing in the PR description: the Jello transactions call setRespectedGameType on the OptimismPortal proxy, not on the dispute game factory. It doesn't affect the docs change.

[palango]

With this line saying Jovian has been activated, line 48 ("Node operators will need to upgrade ... before the activation dates") and the withdrawal steps at lines 39-41 now read as pending. Could they go to past tense? Please keep step 4 (resubmit an invalidated withdrawal proof), which still applies.

[palango]

This phase is finished like the four above it. I'd add the ✅ so it doesn't read as still open.

[palango]

Date checks out (slot 13164544 = 2025-12-03 21:49:11 UTC). Since it's in the past now, "affected Celo Mainnet" here and on the Sepolia line above.

[eddiemessiah]

Previewed in a local live deployment (mint dev, this branch merged with the other open docs fixes, no overlapping files) at desktop 1280px and phone 412px:

• /operate/notices/archive/jello-upgrade
• /operate/notices/archive/l1-fusaka-upgrade
• /operate/notices/archive/jovian-upgrade
• /operate/notices/archive/isthmus-upgrade
• /operate/notices/archive/eigenda-v2-upgrade
• /operate/notices/archive/celo-sepolia-launch
• /operate/specification/upgrades/jello

Each page returns 200 and renders the new text, with no horizontal overflow on phone, no broken images and 0 console errors.

[eddiemessiah]

Thanks both. New head is ad00b7e2 (one commit on top of the original, rebased onto current main, force-pushed with --force-with-lease). Every finding is addressed below.

Lena's findings

#	Finding	Status	Where and new wording
1	"will need to upgrade ... before the activation dates"	FIXED	isthmus-upgrade.mdx:42, jovian-upgrade.mdx:48: "Node operators needed to upgrade to the respective Isthmus/Jovian releases before the activation dates." eigenda-v2-upgrade.mdx:31: "Node operators needed to upgrade the EigenDA proxy to version v1.8.2 before the activation date. The version is backwards compatible with EigenDA v1 and could be updated beforehand." Also :33 "will require to add" is now "requires operators to add".
2	Celo Sepolia notice still announces September 2025 as future	FIXED	celo-sepolia-launch.mdx:6: "Celo Sepolia is the developer testnet that replaced Alfajores when Holesky was sunset in September 2025. The Baklava testnet was also sunset with Holesky, with no replacement." :20: "Sep 30, 2025: Alfajores and Baklava sunset, aligned with the Holesky deprecation" with the check mark. :22: "Node Providers: Alfajores and Baklava are retired; support Celo Sepolia." :35: "Node providers were asked to support both Alfajores and Celo Sepolia in parallel during the early access and transition phases ... Alfajores and Baklava are retired, so only Celo Sepolia needs support." :67 table cell: "Sunset in September 2025". The Sep 30 date is the one the page already carried as the plan; the launch forum post (2025-08-13) gives the same date for the Holesky sunset and end of Alfajores/Baklava support, and nothing on main contradicts it. I did not find a Celo-side record of the actual shutdown day, so if you would rather not assert the day I can change :20 to "September 2025".
3	Transition period phase has no check mark	FIXED	celo-sepolia-launch.mdx:19: "Aug 14, 2025 - Sep 14, 2025: Transition period" now ends with the same check mark as the four phases above it.
4	"We recommend the latest released versions"	FIXED	jello-upgrade.mdx:31 and celo-sepolia-launch.mdx:37: "The versions released for this upgrade / for the Celo Sepolia launch were op-geth v2.1.2, op-node v2.1.0 and eigenda-proxy v1.8.2. For current versions see Network config." The version numbers are the ones each page already listed, kept as links to the release tags; nothing was copied in from #2376. l1-fusaka-upgrade.mdx:30: "No upgrade of the Celo L2 client was required for the L1 Fusaka upgrade. The Celo client versions recommended at the time were ..." plus the same Network config link, since nothing was released for Fusaka itself. jello-upgrade.mdx:29 is also past tense: "Node operators did not need to upgrade the Celo client software for this upgrade."
5	Fusaka page: present and future wording around the past date	FIXED	l1-fusaka-upgrade.mdx:12: "Fusaka activated on Ethereum Mainnet on Wednesday, December 3rd, 2025 at 21:49:11 UTC and on Ethereum Sepolia on Tuesday, October 14th, 2025 at 07:36:00 UTC; it affected Celo Mainnet and Celo Sepolia respectively." :14: "node operators had to provide Fusaka-compatible L1 node access". :17 Note: "The Fusaka upgrade required no changes for Celo users." I took that from the page's own L2 client sentence ("There are no upgrades required for the Celo L2 client"), which is still on the page at :30 in past tense. :26: "L1 nodes had to be updated to a Fusaka-compatible version before the fork. The Ethereum client release list names the compatible versions."

palango's inline comments

Point	Status	Where and new wording
Jovian line 48 and withdrawal steps 39-41 to past tense, keep step 4	FIXED	jovian-upgrade.mdx:39-41: "There was a one-time invalidation of all pending withdrawal proofs created on L1." / "Pending withdrawals had to be completed before the upgrade was executed." / "New withdrawal proofs that would not become executable in time had to be avoided." Step 4 at :42 is unchanged. The Note at :32 is now "All withdrawals that were not finalized before the contract upgrade right before the Jovian hardfork need to be reproven (see step 4 below)."; the "consider waiting ... 7-day window" sentence is dropped because the window is over.
Check mark on the transition-period phase	FIXED	celo-sepolia-launch.mdx:19, see finding 3.
"affected" on both Fusaka date lines	FIXED	Both dates are now in one sentence at l1-fusaka-upgrade.mdx:12 ending "it affected Celo Mainnet and Celo Sepolia respectively".
PR description: setRespectedGameType target	FIXED	The body now reads "setRespectedGameType on the L1 dispute game factory on the L1 OptimismPortal proxy". The rest of the body is as before, so its "What this does NOT do" list describes the first revision; the first three items there are now done by this commit.

Checks on ad00b7e2

• git merge-tree --write-tree origin/main HEAD: clean (the branch is now rebased onto main at 7b8f97c2, so the tree is main plus the six notices).
• mint broken-links --check-redirects, run on Windows on origin/main and on this head, outputs diffed: the only differences are the removed isthmus-upgrade.mdx -> /operate/operators/faq entry from the first commit and three new entries for /operate/operators/network-config (one per page that now links it). On Windows the CLI reports root-relative internal links as broken (the origin/main run already lists the same target six times from existing pages), so those three are the known Windows path issue, not missing pages: the page exists on disk, is in docs.json navigation, and the local preview serves it with 200. I could not get a Linux run on this machine (WSL would not start), so CI is the authoritative broken-links result.
• bash scripts/check-orphans.sh: "No orphan pages found." jq is not installed here, so I ran the script with a small Node stand-in that implements exactly the two jq filters the script uses.
• git grep -n "will need to upgrade\|Please support both Alfajores\|We recommend the latest released versions\|no changes are expected" operate/notices: no matches.
• mint dev --port 3472: all six pages (/operate/notices/archive/{jello-upgrade,l1-fusaka-upgrade,jovian-upgrade,isthmus-upgrade,eigenda-v2-upgrade,celo-sepolia-launch}) return 200 and the served HTML contains the new sentences quoted above; /operate/operators/network-config returns 200.
• Playwright at 1280x900 and 412x915: all twelve loads returned 200 with no console errors and no horizontal document overflow. I have the screenshots and can attach them if useful.
• All six files are still CRLF.

Diff of this commit: 6 files, 22 insertions, 35 deletions, all under operate/notices/archive/.

[GigaHierz]

head ad00b7e · API headRefOid == local refs/pull ref (re-fetched); 2 commits on base 7b8f97c = current origin/main (0 behind); merge-tree against origin/main clean, and clean against #2376 and #2345; mint broken-links --check-redirects and check-orphans.sh pass on the head; CI: Documentation Validation is action_required (waiting for a maintainer to approve the workflow run, not a failure; GitGuardian and StepSecurity pass); rendered all 7 changed pages on port 3622 (HTTP 200, expected text served), Jovian page opened in a browser tab: 0 console errors, no horizontal overflow

Verdict: REQUEST-CHANGES

Thank you for the quick turnaround. Every finding from the first round is fixed, and all the dates still check out against primary sources. What remains is the same kind of issue in the two pages you did not change this round: the "latest release" sections of the Isthmus and Jovian notices, plus a handful of present and future tense lines that the new "has been activated" banners now sit above.

Status of earlier findings

1. Medium, "will need to upgrade ... before the activation dates": fixed in Isthmus (:42), Jovian (:48) and EigenDA v2 (:31), all in past tense now.
2. Medium, Celo Sepolia launch still announcing September 2025 as future: fixed. Intro, phase list, node-provider call to action, the Longevity cell and the "Please support Alfajores" line are all past or retired wording.
3. Low, missing check mark on the transition period: fixed (and the Sep 30 sunset line has one).
4. Low, "We recommend the latest released versions": fixed on the Jello, Fusaka and Celo Sepolia launch pages, which now say what was released at the time and link to Network config. The same pattern is still on Isthmus and Jovian, see finding 1.
5. Low, Fusaka wording: fixed. The date line and the call to action are past tense. The Note "required no changes for Celo users" rests on the page's own L2 client statement; I am comfortable with it.

Findings

1. Medium · operate/notices/archive/jovian-upgrade.mdx:50-56 and isthmus-upgrade.mdx:44-49 · "Update to the latest release" lists versions that are no longer the latest

Jovian lists op-geth v2.2.2, op-node v2.2.1 and eigenda-proxy v2.6.0 under the heading "Update to the latest release"; Isthmus lists op-geth v2.1.0 and op-node v2.1.0 under the same heading. The latest op-node release is celo-v2.2.3 (2026-10-05), and compose, which pins celo-v2.2.3, is what Network config documents once the sibling PR lands. The celo-v2.2.3 release notes say it is required for the Glamsterdam activation on Celo Sepolia on 6 October. op-geth support has ended (see the op-geth notice). This is the same defect you fixed on the Jello, Fusaka and Celo Sepolia pages.

Failure scenario: an operator or an AI assistant reads the Jovian page, sees "latest release" with op-node v2.2.1, and installs it. On Celo Sepolia that version predates Amsterdam L1 header support, so the node stops following the chain once Glamsterdam activates.

Fix shape: rename both headings to something time-bound, for example "Releases for Jovian" and "Releases for Isthmus", rewrite the lead-in in past tense ("The release contained the activation timestamps ..."), and add the same sentence as the other pages: "For current versions see Network config." Keep the version numbers (they are correct for their date).

2. Low · jovian-upgrade.mdx:32, :92, :106, :114-116 and isthmus-upgrade.mdx:55, :61 · remaining present and future tense under the "has been activated" banner

• jovian-upgrade.mdx:32: "withdrawals that were not finalized ... need to be reproven" mixes past with present; the list below it is past tense. Suggest "had to be reproven (see step 4 below)".
• jovian-upgrade.mdx:92: "Node operators not using the Docker Compose setup must re-download these files before upgrading." Suggest "had to re-download these files before upgrading".
• jovian-upgrade.mdx:106 and isthmus-upgrade.mdx:61: "op-node and op-geth will log their configurations at startup" under "Verify Your Configuration". Suggest "logged".
• jovian-upgrade.mdx:114-116: "Nodes will perform a one-time log index rebuild on first start ... Subsequent restarts will not re-index." Suggest "performed ... did not re-index".
• isthmus-upgrade.mdx:55: "operators need to make sure to upgrade the EigenDA proxy to version v1.8.2". Suggest "operators needed to upgrade".

Failure scenario: a reader (or an assistant answering "do I still need to re-download rollup.json or re-prove my withdrawal?") cannot tell whether an action is still open on Celo Sepolia or mainnet, even though both have activated. This is the confusion the PR set out to remove.

Fix shape: one-word tense edits on those six lines. If you would rather keep the "Verify Your Configuration" and "Notes" blocks as timeless operational notes, say so in the PR body and keep the headings clearly historical; otherwise past tense is the consistent choice.

3. Low · operate/specification/upgrades/ice-cream.mdx:20 · the page the EigenDA v2 notice links to still says the sequencer "will use" EigenDA v2 and nodes "must upgrade"

The PR changes jello.mdx in the same directory and puts the Ice Cream notice in past tense, but the spec page ice-cream.mdx:20 ("After the Celo Ice Cream update, the Celo sequencer will use EigenDA v2 ... must upgrade it's EigenDA proxy") is still future tense (Ice Cream activated on mainnet on Sep 10, 2025). It is the only remaining will under specification/upgrades/.

Failure scenario: a reader follows the notice's link to the spec and reads the upgrade as pending.

Fix shape: "the Celo sequencer uses EigenDA v2 ... nodes had to upgrade their EigenDA proxy", and fix the it's typo (its). One line, same concern as this PR; if you prefer to keep this PR to the notices, file a follow-up issue and link it from the PR body.

Verified good

All dates recomputed by me from sources independent of the PR text:

• Jello Celo Sepolia: tx 0x736deb75...b31a in Ethereum Sepolia block 9,565,574, Blockscout 2025-11-05T11:16:00Z, setRespectedGameType to 0x44ae3d41..., which equals OptimismPortalProxyAddress in the Celo Sepolia deployment-l1.json. 1762341360 converts to Wed 2025-11-05 11:16:00 UTC. Matches jello.mdx and the notice (the old "Oct 5" was wrong).
• Jello mainnet: tx 0x5fb3f222...c44b in Ethereum block 23,981,692, Blockscout 2025-12-10T10:24:11Z, setRespectedGameType to 0xc5c5D157..., which equals OptimismPortalProxy in the mainnet deployment-l1.json. The old 10:24:22 does not match the block; 10:24:11 is right in both the spec table and the notice.
• Fusaka mainnet: slot 13,164,544 x 12 s + genesis 1606824023 = 1764798551 = Wed 2025-12-03 21:49:11 UTC. Fusaka Sepolia: 1760427360 = Tue 2025-10-14 07:36:00 UTC. Weekdays in the page are correct.
• Jovian: rollup.json jovian_time 1774958788 = Tue 2026-03-31 12:06:28 UTC (mainnet) and 1773749037 = Tue 2026-03-17 12:03:57 UTC (Celo Sepolia); both match the notice.
• Isthmus: mainnet isthmus_time 1752073200 = Wed 2025-07-09 15:00:00 UTC; the retired Baklava (1749654000, Wed 2025-06-11 15:00 UTC) and Alfajores (1750863600, Wed 2025-06-25 15:00 UTC) timestamps convert to the stated times. The Celo Sepolia rollup config has isthmus_time 0 (genesis).
• Ice Cream dates (Jul 30, Aug 20, Sep 10, 2025) were not changed by the PR and not re-derived; they match ice-cream.mdx.
• No page the PR changed contradicts its spec page; no frontmatter changed; headings unchanged; the new link /operate/operators/network-config exists. The removed FAQ link in Isthmus pointed to an entry that no longer exists. The diff adds no anchors (the "step 4 below" reference is plain text).
• Not checked: the Alfajores and Baklava chains are retired, so their activation blocks could not be re-read; I used rollup.json for both networks, not a Celo-side config.json.

What the PR got right

• Using the L1 block time for Jello (an L1 contract change) is the right source, and the evidence reproduces exactly; it caught a wrong month, an 11-second drift, a typo and a dead FAQ link in one pass.
• The follow-up commit addressed every earlier finding, including the Celo Sepolia launch page and the Fusaka page, without widening scope.
• A tight, reviewable diff, and a measured choice on the Alfajores wording.

Merge-order hazards

• #2376: git merge-tree of the two heads is clean. The notices now say "For current versions see Network config", so merge #2376 first or together; without it that page still lists op-node v2.1.0 and eigenda-proxy v1.8.2.
• #2345 (draft) touches only operate/operators/*; clean against this PR.
• #2346 (Alfajores references) may touch celo-sepolia-launch.mdx; if it does, whoever merges second will conflict on the lines fixed here.
• Do not copy version numbers into the notices from either PR; link to Network config instead.

[palango]

Thanks for the follow-up. Everything from my last review is in: lines 39-41 and 48 of the Jovian notice are past tense with step 4 kept, the transition period has its check mark, both Fusaka dates say "affected", and the Alfajores call to action is gone. Leaving the Jovian note at line 32 in the present tense is right, since step 4 still applies to anyone who never re-proved.

The new commit adds one factual problem. The Celo Sepolia notice now says Holesky "was sunset in September 2025", but September was only the plan. Fusaka went live on Holesky in early October 2025, and the EF's shutdown announcement (https://blog.ethereum.org/2025/09/01/holesky-shutdown-announcement) put the shutdown two weeks after Fusaka finalized there. Since we have no record of the day Alfajores and Baklava stopped, I'd stick to what the forum post said, that 30 September was the end of support. Suggestions are inline.

Two smaller ones are also inline: the Jello page now says versions were "released for this upgrade" right after saying no upgrade was needed, and "requires operators to add" on the EigenDA v2 page reads as a current instruction for flags the Jovian upgrade later removed.

I agree with GigaHierz's point about the "Update to the latest release" sections on Isthmus and Jovian. And the PR body's "does NOT do" list and the judgement call about the check mark still describe the first revision, so please refresh them before merge.

[palango]

Holesky wasn't sunset in September 2025. Fusaka only reached it in early October, and the EF shut it down a couple of weeks after that (https://blog.ethereum.org/2025/09/01/holesky-shutdown-announcement). How about: "Celo Sepolia is the developer testnet that replaced Alfajores after Ethereum deprecated Holesky in 2025. Baklava was retired at the same time, with no replacement."

[eddiemessiah]

Fixed in f7b2da5 with your wording: "Celo Sepolia is the developer testnet that replaced Alfajores after Ethereum deprecated Holesky in 2025. Baklava was retired at the same time, with no replacement." The EF post confirms the shutdown was tied to two weeks after Fusaka finalized on Holešky, not to September.

[palango]

Same issue as line 6: the Holesky shutdown came later, and we have no record of the day Alfajores stopped. The forum post called 30 September the end of support, so I'd write "Sep 30, 2025: End of support for Alfajores and Baklava ✅". Line 67 could then just say "Retired".

[eddiemessiah]

Fixed in f7b2da5: "Sep 30, 2025: End of support for Alfajores and Baklava ✅", and the Longevity cell on line 67 now says "Retired".

[palango]

Line 29 says no upgrade was needed, so nothing was released for Jello. These versions predate it (op-node celo-v2.1.0 is from June 2025, op-geth celo-v2.1.2 from August). The Fusaka page's wording fits better: "The Celo client versions recommended at the time were ..."

[eddiemessiah]

Fixed in f7b2da5: "The Celo client versions recommended at the time were ...". Confirmed the tags predate Jello: op-node celo-v2.1.0 was published 2025-06-25 and op-geth celo-v2.1.2 on 2025-08-13.

[palango]

In the present tense this reads as a current instruction, but the Jovian upgrade tells operators to remove most of these flags (jovian-upgrade.mdx lines 72-82). "required operators to add" keeps it historical.

[eddiemessiah]

Fixed in f7b2da5: "required operators to add", so it reads as historical next to the Jovian flag removals.

[eddiemessiah]

Thanks Lena, all three are fixed in 1f61865.

• F1, release lists (Jovian, Isthmus): FIXED. The heading is now "Releases for Jovian" / "Releases for Isthmus", the lead-in reads "The release contained the activation timestamps …", and it links to Network config for current versions. The version numbers are kept. Nothing in the repo linked to the old #update-to-the-latest-release anchor.
• F2, tense (Jovian, Isthmus): FIXED. The changed lines are "had to be reproven", "had to re-download", "logged their configurations" (on both pages), "performed a one-time log index rebuild … did not re-index" and "operators needed to upgrade the EigenDA proxy to version v1.8.2".
• F3, Ice Cream spec: FIXED. It now reads "the Celo sequencer uses EigenDA v2 … nodes following the network had to upgrade their EigenDA proxy", which also fixes "it's".

I previewed the three pages locally at 1280px and 412px. There was no overflow and no console errors. Because these pages now link to Network config for current versions, this PR should merge after #2376 or together with it.

[palango]

The new commit looks good to me. The Isthmus and Jovian release sections read as historical now, and the Ice Cream spec line is fixed without touching any headings.

My four inline comments from the last round are still open, though, and the files they're on haven't changed. The one I care about most is the Holesky date on the Celo Sepolia notice (lines 6 and 20). The EF shutdown post (https://blog.ethereum.org/2025/09/01/holesky-shutdown-announcement) ties the shutdown to two weeks after Fusaka finalized on Holesky, not to September. The forum post only gives 30 September as the end of support, so I'd say that and leave Holesky's date out. The other two are smaller. The Jello page still says versions were "released for this upgrade" right after saying no upgrade was needed, and those tags predate Jello. "Requires operators to add" on the EigenDA v2 page reads as current advice for flags the Jovian notice tells people to remove. If you'd rather keep that last one, a reply on the thread is fine.

Documentation Validation hasn't run on this head or the previous one; it's waiting for a maintainer to approve the workflow. mint broken-links passes locally. And since #2376 is still open, Network config still lists op-node celo-v2.1.0, so this should merge after #2376.

[eddiemessiah]

@palango thanks, sorry I missed your four inline comments last round. All four are fixed in f7b2da5 and answered on their threads: the Holesky date is out and 30 September is given as the end of support for Alfajores and Baklava, the Jello versions now read as recommended at the time, and the EigenDA v2 flags line is past tense. Agreed on the merge order: after #2376 or together with it.

[GigaHierz]

Re-reviewed the current head: the findings from the earlier review are resolved, the diff merges cleanly into current main, and the changed claims still check out.