docs(archive): review state archive performance work

Draft
BL
19d ago
+104548-316

What does this PR do?

Opens the current State Archive implementation and performance evidence for code-difference review only.

Why are these changes required?

The branch contains the current State Archive persistence, recovery, serving, ARM/AMD validation, and performance-tracking changes that need review before any merge decision.

This PR has been tested by:

  • Existing branch-level validation and recorded State Archive harness evidence
  • No new product change is introduced by opening this Draft PR

Follow up

Review only. Keep this PR in Draft; do not merge or enable auto-merge.

Extra details

This PR is intentionally created in the origin fork bladehan1/java-tron, not tronprotocol/java-tron. No upstream repository or runtime deployment is changed.

Summary by cubic

Review-only PR for the State Archive branch. It adds an opt-in archive for durable reverse history, recovery, serving, historical reads, and path-state roots; with archive features disabled, the existing current-state path keeps its current behavior.

State Archive

  • Captures canonical block snapshots and persists bounded history body, index, and commit records, partitioned per store type, with backpressure, reorg handling, and restart-safe recovery.
  • Adds exact-27 startup admission with diagnostics, native LevelDB/RocksDB recovery, immutable serving generations, engine identity checks, and fail-closed drift detection.
  • Supports request-owned historical point and range reads with limits and cancellation; historical VM execution rejects SELFDESTRUCT, CREATE, and CREATE2.
  • Adds path-state rebuild, catch-up, transitions, reorg rewind, P66 migration, compaction, runtime status, and non-consensus block metadata.
  • Adds offline block replay, file tools, and resumable HTTP export.

Performance Evidence

  • Records ARM/AMD HLT-002 observations and JDK17/rocksdb 9.7.4 comparisons while retaining JDK8 controls.
  • Includes layered trie deltas, owned snapshot heads, parallel bounded rebuilds with progress, and reduced cross-store reads through lazy migration.

Keep this PR in Draft; do not merge or auto-merge.