docs(changelog): withdraw the "q8 stays a win over Kuzu" claim (#1825)

[lmeyerov]

Master carries a benchmark claim that my own re-measurement refutes. Correcting it in place rather than leaving it to round-trip as fact.

#1823's CHANGELOG entry closes with:

The cells this lane is not called for are unchanged: q5 and q8 are ties at both scales in both runs, and q8 stays a win over Kuzu.

The first half is a real measurement of what that lane does, and it stands. The second half is a board-position claim, and it does not survive re-measurement.

Why it fails

q8's headline number is a cross-call memoization artifact. The two-hop equal-domain degree counts are cached onto the caller's Plottable, keyed by id(), so the figure quoted on the board is a warm repeat call:

arm	20k	100k
warm (memo hit)	2.02 ms	5.09 ms
cold (memo delattr'd)	13.18 ms	49.68 ms
fresh Plottable	14.60 ms	53.02 ms
bind_only control	0.02 ms	0.02 ms

The bind_only arm is the one that makes this conclusive: re-binding costs 0.02–0.04 ms, so the gap between warm and cold is the memo, not setup. A one-shot q8 loses to embedded Kuzu by 3.1–5.2× at 20k and 2.8–6.0× at 100k.

The mechanism is itself filed as #1825: keying a cache by id() on an object the caller still owns also returns a stale answer after an in-place edge or node mutation, on all engines. So the cell is not merely optimistic — what produces it is a bug.

Scope

Reconciled, not deleted, per the standing rule that reconciling beats deleting:

• the withdrawn sentence is quoted in place, so the correction is traceable from the original claim;
• the measurement perf(gfql): fuse the polars single-hop grouped aggregate into one lazy plan #1823 actually reports — that its lane leaves q8 unchanged — is explicitly preserved;
• no other entry, number, or line is touched. CHANGELOG-only, one paragraph.

Related

Two sibling instances of the same claim were caught before merge and are already handled: #1816 carried it in its own entry (removed, now points at #1825), and #1818 ships graphbench.{20k,100k}.q8.polars_vs_kuzu as board_quotable: true with an empty disclosures list — that one is deferred pending a regenerated board.