feat: blockstm remove incarnation cache

[Eric-Warehime]

Description

Removes incarnation cache from blockstm.

Reasoning being that we've just removed the usage of incarnation cache for caching signature verification checks. This was removed in #25912 due to caching causing indeterminism bugs.

After this there are no current usages of the cache, and in general the cache promotes patterns which are likely to cause indeterminism and app hashes.

[greptile-apps]

Greptile Summary

This PR removes the incarnation cache from BlockSTM, eliminating a shared mutable cache that was passed between transaction incarnations during parallel block execution. The cache was previously used to deduplicate signature verification, but its last consumer was removed in #25912 due to indeterminism bugs, leaving the mechanism with no active users.

Key changes:

• incarnationCache map[string]any parameter removed from DeliverTxFunc, RunTx, and deliverTx across baseapp and internal/blockstm
• incarnationCache field and all associated methods (GetIncarnationCache, SetIncarnationCache, WithIncarnationCache, IncarnationCache) removed from sdk.Context
• Cache type alias and cache-aware Tx function signature removed from the BlockSTM mock/test infrastructure
• TestSTMRunner_Run_IncarnationCache test deleted as it is no longer meaningful

Minor issues found:

• The first line of the StartSpan GoDoc comment in types/context.go was accidentally deleted during the removal
• The unused SigVerificationResultCacheKey constant in internal/blockstm/mock_block.go should be cleaned up

Confidence Score: 4/5

• Safe to merge — the removal is consistent across all call sites and the deleted functionality had no remaining consumers.
• The change is a clean, mechanical removal of a parameter and associated struct field across 13 files. All call sites are updated consistently, no logic was silently dropped beyond the explicitly-intended cache. Two minor style/housekeeping issues remain: an accidentally deleted StartSpan doc comment line and an orphaned constant. These don't affect functionality but should be addressed for code quality.
• types/context.go (truncated GoDoc) and internal/blockstm/mock_block.go (orphaned constant).

Sequence Diagram

sequenceDiagram
    participant STMRunner
    participant deliverTx as DeliverTxFunc
    participant RunTx as BaseApp.RunTx
    participant Context as sdk.Context

    Note over STMRunner: Before this PR
    STMRunner->>STMRunner: init incarnationCache[]
    STMRunner->>deliverTx: deliverTx(tx, memTx, ms, idx, cache)
    deliverTx->>RunTx: RunTx(mode, tx, memTx, idx, ms, cache)
    RunTx->>Context: ctx.WithIncarnationCache(cache)
    Context-->>RunTx: ctx with cache
    RunTx-->>deliverTx: result

    Note over STMRunner: After this PR
    STMRunner->>deliverTx: deliverTx(tx, memTx, ms, idx)
    deliverTx->>RunTx: RunTx(mode, tx, memTx, idx, ms)
    RunTx-->>deliverTx: result

Loading

Comments Outside Diff (1)

1. internal/blockstm/mock_block.go, line 14 (link)

   SigVerificationResultCacheKey is no longer referenced anywhere in the codebase following the removal of incarnation cache logic. This constant should be deleted.

_{Last reviewed commit: d6b9594}

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.78%. Comparing base (ead3b81) to head (df0caab).
⚠️ Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
baseapp/abci.go	50.00%	1 Missing ⚠️
baseapp/test_helpers.go	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #26048      +/-   ##
==========================================
- Coverage   59.88%   59.78%   -0.11%     
==========================================
  Files         982      966      -16     
  Lines       65193    64267     -926     
==========================================
- Hits        39038    38419     -619     
+ Misses      26155    25848     -307     

Files with missing lines	Coverage Δ
baseapp/baseapp.go	85.06% <100.00%> (-0.06%)	⬇️
baseapp/genesis.go	80.00% <100.00%> (ø)
baseapp/txnrunner/default.go	100.00% <100.00%> (ø)
internal/blockstm/txnrunner.go	79.06% <100.00%> (-2.19%)	⬇️
types/abci.go	100.00% <ø> (ø)
types/context.go	91.70% <ø> (+4.33%)	⬆️
baseapp/abci.go	85.94% <50.00%> (ø)
baseapp/test_helpers.go	66.66% <75.00%> (ø)

... and 21 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mmsqe]

we met nondeterministic isssue in ibc test before, but thinking it's ok to keep the interface to support eth signature-result cache later.

[yihuang]

EVM tx signature verification is one use case, if we don't do it in with incarnation cache, we might want to have a special pipeline for signature verification, because in block-stm, it'll make it heavier to re-execute aborted transactions.

[Eric-Warehime]

We'll keep this here for now--caching signature validation should be stateless (in the sense of not requiring store access) and so should be safe for caching.