[Verdenroz]

What changed

With [gate] walltime = "advisory", gate --against-ref now measures each side once, so it runs two rounds instead of four. Rounds 3 and 4 skip the gating counters and only add a second walltime sample per side. Advisory walltime never fails, so that second sample can't change the verdict. On a real project's PR gate the measure phase took 5.2 of 15.6 minutes over four rounds. Now that the base build is seeded, measuring is the longest phase of that gate.

• Passed walltime_advisory from gate and gate accept through resolve into measure_ref_interleaved. Under advisory the timing-only rounds 3 and 4 are not run.
• Changed the round line to show the real total: gate: round 1/2 (head) under advisory, round n/4 otherwise.
• Kept combine_rounds unchanged. Its caller uses each side's first round directly when the timing-only rounds were skipped.
• Documented in docs/gating.md that advisory mode measures each side once, head first. A baseline it saves, and a run it stores in the run cache, then holds one walltime sample per item instead of the minimum of two.

Instructions, Ir, allocations, size and a bench's own p99 claim come from the first round, so they gate exactly as before. The default, walltime = "hard", still runs four rounds.

Why

The measure phase is now the largest part of an adopter's PR gate. Half of it bought a walltime sample that advisory mode reports as SOFT at most.

How was this tested

• Ran a live gate on a real project's clone (--against-ref HEAD~5, one item, --no-reuse-base --no-item-cache). With walltime = "advisory" it printed round 1/2 (head) and round 2/2 (base) and passed with exit 0. Without the setting it printed rounds 1/4 to 4/4 and passed with exit 0. With one filtered item each round printed 0s, so this run shows the round count, not the time saved.
• make check: exit 0, 449, 6 and 4 tests passed (test: 1 of 10 packages selected, 3 binaries run).
• cargo clippy -p cargo-soothfast --all-targets -- -D warnings: clean.
• No unit test covers the skip itself. It lives inside the merge-base worktree flow, which the live run exercised.
• make gate BASE=master was not run. No measured code changed.

Checklist

• make check passes (fmt, clippy -D warnings, cargo test --workspace)
• make gate BASE=master passes, or any intentional cost change is explained above
• Tests added/updated for new behavior (not required for docs/CI-only changes)
• Public API items have /// doc comments
• No new dependency, or its justification is included above (see Dependency Policy)
• Docs (README.md, docs/, soothfast:bind/soothfast:claim markers) updated if behavior changed

[soothfast-bot]

soothfast gate: soothfast-registry

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 6 dependency units (24.2 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.13% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_registry::bench_fnv1a walltime_median_ns 81531.3 -> 81531.3 (+0.0%)
ok    soothfast_registry::bench_fnv1a allocs 0 -> 0 (allowed <= 0)
ok    soothfast_registry::bench_fnv1a alloc_bytes 0 -> 0 (allowed <= 0)
ok    soothfast_registry::bench_yield_chain walltime_median_ns 139.4 -> 139.4 (+0.0%)
ok    soothfast_registry::bench_yield_chain allocs 1 -> 1 (allowed <= 1)
ok    soothfast_registry::bench_yield_chain alloc_bytes 24 -> 24 (allowed <= 25)
ok    soothfast_registry::bench_yield_chain polls 9 -> 9 (allowed <= 9)
ok    soothfast_registry::bench_yield_chain wakes 8 -> 8 (allowed <= 8)
ok    soothfast_registry::bench_fnv1a assert alloc: allocs 0 <= 0
ok    soothfast_registry::bench_fnv1a assert complexity: claimed O(n); growth drift x1.01 over sizes [1024, 8192, 65536] (limit x2.5)
ok    soothfast_registry::bench_yield_chain assert alloc: allocs 1 <= 1
gate: passed (2 item(s))

[soothfast-bot]

soothfast gate: soothfast-measure

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 6 dependency units (24.2 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.20% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_measure::bench_summarize walltime_median_ns 608763.7 -> 608763.7 (+0.0%)
ok    soothfast_measure::bench_summarize allocs 4 -> 4 (allowed <= 4)
ok    soothfast_measure::bench_summarize alloc_bytes 524288 -> 524288 (allowed <= 550502)
ok    soothfast_measure::bench_sweep_evaluate walltime_median_ns 24.5 -> 24.5 (+0.0%)
ok    soothfast_measure::bench_sweep_evaluate allocs 0 -> 0 (allowed <= 0)
ok    soothfast_measure::bench_sweep_evaluate alloc_bytes 0 -> 0 (allowed <= 0)
ok    soothfast_measure::bench_summarize assert alloc: allocs 4 <= 8
ok    soothfast_measure::bench_summarize assert complexity: claimed O(n log n); growth drift x1.32 over sizes [1024, 4096, 16384] (limit x2.5)
ok    soothfast_measure::bench_sweep_evaluate assert alloc: allocs 0 <= 0
gate: passed (2 item(s))

[soothfast-bot]

soothfast gate: soothfast-docs

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 17 dependency units (44.9 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.11% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_docs::bench_claim_parse walltime_median_ns 213.8 -> 213.8 (+0.0%)
ok    soothfast_docs::bench_claim_parse allocs 6 -> 6 (allowed <= 6)
ok    soothfast_docs::bench_claim_parse alloc_bytes 170 -> 170 (allowed <= 178)
ok    soothfast_docs::bench_markdown_scan walltime_median_ns 413968.0 -> 413968.0 (+0.0%)
ok    soothfast_docs::bench_markdown_scan allocs 4114 -> 4114 (allowed <= 4319)
ok    soothfast_docs::bench_markdown_scan alloc_bytes 338400 -> 338400 (allowed <= 355320)
ok    soothfast_docs::bench_claim_parse assert p99: p99 242ns <= 50000ns
ok    soothfast_docs::bench_markdown_scan assert complexity: claimed O(n); growth drift x0.95 over sizes [512, 2048, 8192] (limit x2.5)
gate: passed (2 item(s))

[soothfast-bot]

soothfast gate: soothfast-report

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 17 dependency units (44.9 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.05% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_report::bench_llms_render walltime_median_ns 309078.2 -> 309078.2 (+0.0%)
ok    soothfast_report::bench_llms_render allocs 7191 -> 7191 (allowed <= 7550)
ok    soothfast_report::bench_llms_render alloc_bytes 1197537 -> 1197537 (allowed <= 1257413)
ok    soothfast_report::bench_perf_table walltime_median_ns 821631.0 -> 821631.0 (+0.0%)
ok    soothfast_report::bench_perf_table allocs 8204 -> 8204 (allowed <= 8614)
ok    soothfast_report::bench_perf_table alloc_bytes 465747 -> 465747 (allowed <= 489034)
ok    soothfast_report::bench_llms_render assert complexity: claimed O(n); growth drift x0.99 over sizes [64, 256, 1024] (limit x2.5)
ok    soothfast_report::bench_perf_table assert complexity: claimed O(n); growth drift x0.92 over sizes [64, 256, 1024] (limit x2.5)
gate: passed (2 item(s))

[soothfast-bot]

soothfast gate: soothfast-site

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 17 dependency units (44.9 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.06% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_site::bench_highlight walltime_median_ns 2845675.0 -> 2845675.0 (+0.0%)
ok    soothfast_site::bench_highlight allocs 88069 -> 88069 (allowed <= 92472)
ok    soothfast_site::bench_highlight alloc_bytes 2079336 -> 2079336 (allowed <= 2183302)
ok    soothfast_site::bench_md_render walltime_median_ns 3345469.0 -> 3345469.0 (+0.0%)
ok    soothfast_site::bench_md_render allocs 68118 -> 68118 (allowed <= 71523)
ok    soothfast_site::bench_md_render alloc_bytes 3502526 -> 3502526 (allowed <= 3677652)
ok    soothfast_site::bench_highlight assert complexity: claimed O(n); growth drift x1.22 over sizes [128, 512, 2048] (limit x2.5)
ok    soothfast_site::bench_md_render assert complexity: claimed O(n); growth drift x1.11 over sizes [256, 1024, 4096] (limit x2.5)
gate: passed (2 item(s))

[soothfast-bot]

soothfast gate: soothfast-spec

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 24 dependency units (54.7 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.12% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_spec::bench_openapi_diff walltime_median_ns 3311996.0 -> 3311996.0 (+0.0%)
ok    soothfast_spec::bench_openapi_diff allocs 77722 -> 77722 (allowed <= 81608)
ok    soothfast_spec::bench_openapi_diff alloc_bytes 10815487 -> 10815487 (allowed <= 11356261)
ok    soothfast_spec::bench_openapi_document walltime_median_ns 1241679.0 -> 1241679.0 (+0.0%)
ok    soothfast_spec::bench_openapi_document allocs 30152 -> 30152 (allowed <= 31659)
ok    soothfast_spec::bench_openapi_document alloc_bytes 4715065 -> 4715065 (allowed <= 4950818)
ok    soothfast_spec::bench_serialize_yaml walltime_median_ns 4215125.0 -> 4215125.0 (+0.0%)
ok    soothfast_spec::bench_serialize_yaml allocs 74277 -> 74277 (allowed <= 77990)
ok    soothfast_spec::bench_serialize_yaml alloc_bytes 8533615 -> 8533615 (allowed <= 8960295)
ok    soothfast_spec::bench_openapi_diff assert complexity: claimed O(n); growth drift x1.03 over sizes [16, 64, 256] (limit x2.5)
ok    soothfast_spec::bench_openapi_document assert complexity: claimed O(n); growth drift x1.08 over sizes [16, 64, 256] (limit x2.5)
ok    soothfast_spec::bench_serialize_yaml assert complexity: claimed O(n); growth drift x0.99 over sizes [16, 64, 256] (limit x2.5)
gate: passed (3 item(s))

[soothfast-bot]

soothfast gate: soothfast-sdk

gate: run cache keyed on callgrind (guest 80eb05cd5d172b67 glibc 2.43 8fc174992c9ed915 valgrind-3.26.0)
gate: measuring merge-base of origin/master in worktree (interleaved rounds)
gate: seeded 24 dependency units (54.7 MiB) from HEAD's target dir in 0.1s
gate: bench binaries identical (code and data match) — no measurable change possible
gate: gating backend = walltime
gate: build=26cf1b noise_floor=0.11% thresholds: instructions +5% ir +5% walltime +10.0% alloc/size +5% polls/wakes +5%
ok    soothfast_sdk::bench_emit_typescript walltime_median_ns 2289496.0 -> 2289496.0 (+0.0%)
ok    soothfast_sdk::bench_emit_typescript allocs 42930 -> 42930 (allowed <= 45076)
ok    soothfast_sdk::bench_emit_typescript alloc_bytes 2938940 -> 2938940 (allowed <= 3085887)
ok    soothfast_sdk::bench_lower walltime_median_ns 1296542.0 -> 1296542.0 (+0.0%)
ok    soothfast_sdk::bench_lower allocs 23907 -> 23907 (allowed <= 25102)
ok    soothfast_sdk::bench_lower alloc_bytes 2064788 -> 2064788 (allowed <= 2168027)
ok    soothfast_sdk::bench_emit_typescript assert complexity: claimed O(n); growth drift x1.00 over sizes [16, 64, 256] (limit x2.5)
ok    soothfast_sdk::bench_lower assert complexity: claimed O(n); growth drift x1.05 over sizes [16, 64, 256] (limit x2.5)
gate: passed (2 item(s))