docs(perf): distinguish implemented from merged

Author: metaphorics

docs/solutions/performance-issues/processing-bound-sync-performance-evolution.md

@@ -520,12 +520,12 @@ The reducer also matched the scalar root at every Merkle level for all blocks 0
 
 The 2.831s value is an overlap ceiling, not removable work. Moving the append before apply admission or into the event loop must still preserve side-branch bodies for heavier-chain reorgs, aggregate per-file height metadata across a batch, and define recovery after an append succeeds but index publication fails. It would also put synchronous storage I/O on the event loop. The design review rejected that lifecycle cost for a bounded 2.831s ceiling. Do not delete persistence or weaken reorg availability to improve the benchmark.
 
-**The measured AVX2 result closes the only remaining stage lever that passed its gate.**
+**PR #30 implements the only remaining stage lever that passed its gate.**
 
 | Item | Verdict |
 |---|---|
 | `utxo_commit` | real work; the shard fan-out is not taken at this block size |
-| `block_rules` | **closed and landed**; prepared-txid reuse plus AVX2 multi-buffer SHA256d64 |
+| `block_rules` | **implemented in PR #30**; prepared-txid reuse plus AVX2 multi-buffer SHA256d64; closes when the PR merges |
 | `block_body_persist` | rejected; 2.831s overlap ceiling does not justify the reorg and event-loop failure surface |
 
 The final matched comparison interleaved three candidate runs with three Bitcoin Core runs on CPU set `0-31`, with a 30-second cooldown and the same local 0→150k corpus: