Revert persisting the EVM-only execution cursor with FlatKV state

[masih]

This reverts #4190, which persisted the EVM-only execution cursor (height, app hash, parent hash, gas limit) as an evmonly/cursor changeset alongside each block's FlatKV state and taught NewEVMOnlyApplication to reload it so Info() reported the durable height after a restart. The app returns to keeping its position in the in-memory evmOnlyState, advanced by Commit, and the BlockChangeSetEncoder hook on the executor is removed. Consequently a restarted EVM-only node again reports height 0 from Info(), so #4169 is reopened by this change.

Three commits landed on top of #4190 in evmonlyapp (#4183, #4194, #4205) and conflicted with the automatic revert. Their behaviour is preserved: checkedSenders, EvmChainConfig, EvmBaseFee and EvmCall are re-applied onto the pre-#4190 struct, with EvmCall reading committedHeight, gasLimit, parentHash and the new lastBlockTime field instead of the removed cursor. The chained SHA-256 app hash is unchanged, and the existing evmonlyapp, proxy, giga/evmonly, node and p2p tests pass.

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[github-actions]

The latest Buf updates on your PR. Results from workflow Buf / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
✅ passed	✅ passed	✅ passed	✅ passed	Sep 16, 2026, 8:10 PM

[codecov]

Codecov Report

❌ Patch coverage is 89.09091% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 65.52%. Comparing base (7e3ca24) to head (bcb74c7).
⚠️ Report is 18 commits behind head on giga-1.

Files with missing lines	Patch %	Lines
sei-tendermint/internal/evmonlyapp/app.go	88.88%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           giga-1    #4212      +/-   ##
==========================================
- Coverage   65.55%   65.52%   -0.03%     
==========================================
  Files        2081     2077       -4     
  Lines      157460   157042     -418     
==========================================
- Hits       103222   102909     -313     
+ Misses      54097    53992     -105     
  Partials      141      141              

Flag	Coverage Δ
sei-chain-pr	83.53% <89.09%> (?)
sei-db	?

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
giga/evmonly/executor.go	91.28% <ø> (+0.07%)	⬆️
giga/evmonly/giga_store.go	93.33% <ø> (+0.22%)	⬆️
sei-tendermint/internal/p2p/giga_router_common.go	78.35% <ø> (+0.15%)	⬆️
sei-tendermint/node/public.go	82.47% <100.00%> (ø)
sei-tendermint/internal/evmonlyapp/app.go	82.97% <88.88%> (+1.20%)	⬆️

... and 7 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[cursor]

PR Summary

High Risk
Restart and crash recovery for EVM-only Autobahn nodes no longer align ABCI height/app hash with durable Giga state, which can re-run InitChain or mis-handshake until #4169-style persistence is restored.

Overview
Reverts durable EVM-only chain position in FlatKV: the BlockChangeSetEncoder hook and WithBlockChangeSetEncoder are removed from giga/evmonly, so commits only carry the FlatKV state encoder’s changesets.

evmonlyapp drops cursor.go and storage load on startup. NewEVMOnlyApplication always returns an app with empty in-memory evmOnlyState (height, app hash, parent hash, pending block) advanced by Commit after FinalizeBlock stages evmOnlyPending. Info() and EvmCall read that memory, not disk—so a restarted process reports height 0 until InitChain runs again, even when Giga state on disk is ahead. Init-chain seeding is simplified (no “already seeded at initialHeight−1” resume path). Restart/resume tests are deleted; node no longer handles a restore error from app construction.

Docs and giga router comments no longer say apps must derive Info() from storage or refuse InitChain when blocks already exist.

^{Reviewed by Cursor Bugbot for commit bcb74c7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[seidroid]

A clean, complete revert of #4190: no dangling references to the removed cursor remain, and the three commits that landed on top (checkedSenders, EvmChainConfig, EvmBaseFee, EvmCall) are faithfully re-applied onto the pre-#4190 struct. The only finding is the restart regression the description itself acknowledges (#4169 reopened), which is deliberate but leaves no test recording the restored behaviour.

Findings: 0 blocking | 2 non-blocking | 1 posted inline

Blockers

• None at the file/PR level.

Non-blocking

• [suggestion] The three restart tests and cursor_test.go are removed along with the feature, which is correct, but nothing now pins the restored behaviour. reopenEVMOnlyTestApp survives and is only used for its storage side effect (_, storage = reopenEVMOnlyTestApp(...) at app_test.go:135, so its *evmOnlyApplication return and the require.True(t, ok) assertion are dead). Reusing it for a short characterization test — reopen after committing a block, assert Info().LastBlockHeight == 0 — would record the reopened #4169 behaviour in the tree, so re-landing cursor persistence flips a visible assertion rather than adding one from scratch.
• 1 suggestion(s)/nit(s) flagged inline on specific lines.

[seidroid]

[suggestion] Dropping the constructor's error return also drops the only path that reloaded a durable position, so a restarted node starts at committedHeight == 0 with a state store that still holds every prior block. The consequence is worse than Info() reporting 0: gigaRouterCommon.runExecute branches on last == 0 and calls InitChain, seedInitialStateVersion returns early for InitialHeight == 1 without inspecting the store, and block 1 is then re-executed against state whose nonces and balances have already advanced — the executor refuses it (or CommitStateChanges(1, ...) fails against a store at a higher version), and runExecute returns, so the node does not come up at all.

This is the #4169 regression the description acknowledges reopening, and the scope is the Autobahn load-test application rather than production (TODO(evmonly-production) still gates real traffic), so I read it as an accepted cost of the revert rather than something to fix here. Worth confirming that no load-test environment currently relies on restarting a node over existing EVM-only storage, since those will now fail to start rather than degrade.