[T4rk1n]

Fork PRs get a read-only token, so the benchmarks workflow failed when it tried to post the sticky results comment.

• Write benchmarks/summary.md to the job summary on every run. The benchmark-results artifact is still uploaded.
• Only post the PR comment when the PR branch is in this repo.
• continue-on-error on the comment step, since dependabot PRs also get a read-only token.
• The hard-regression message now points to the job summary.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[github-actions]

Dash performance benchmarks

✅ all within thresholds

	scenario	metric	p90 (ms)	median	growth	baseline p90	note
✅	callback_chain	chain_ms	414.6	402.8	0.93x	499.1
✅	callback_chain	graph_ms	2.4	2.4	1.0x	2.4
✅	callback_fanout	fanout_ms	85.8	76.5	0.87x	92.5
✅	deep_nesting	render_ms	53.3	51.5	0.9x	56.8
✅	full_children_replace	replace_ms	5158.4	1948.1	18.04x	4697.1
✅	initial_render_large	render_ms	595.4	577.6	0.99x	694.4
✅	initial_render_small	render_ms	126.1	102.4	1.07x	104.0
✅	patch_append_nested	append_ms	123.2	83.3	2.33x	192.3
✅	patch_append_toplevel	append_ms	104.9	76.4	2.03x	140.2
✅	patch_scalar_update_large	update_ms	138.4	125.3	0.89x	202.9
✅	wildcard_all_resolve	wildcard_ms	301.6	286.1	0.93x	313.6
✅	wildcard_all_resolve	graph_ms	1.1	1.1	1.0x	1.3

growth = late-third / early-third per-op time; ~1 is flat, a large value means the per-op cost scales with accumulated state.

machine scale vs baseline: 1.07x - divided out of the baseline ratios so they compare like for like (the absolute warn/fail ceilings are left un-scaled); calibrated on initial_render_small.