GitHub mutation-budget alternatives — 2026-08-10

Verdict

INCOMPLETE. The local packet executes fak’s estimator and a direct-call/no-reserve baseline only. It does not claim Octokit, gh api, or Envoy results. Issue #6136 remains open for independently captured real-client witnesses.

Same-workload contract

Every arm receives twelve remaining GitHub API calls, a five-call reserve, the same reset time, and one plan containing eight closes, five comments, and two fetches. The correct result is to hold the 15-call plan before mutation while retaining separate mutation/fetch accounting.

Arm	Class	Local availability	Honest result
fak native mutation reserve	native	yes	correct HOLD; 15 total, 13 mutation, 2 fetch calls
direct API calls without reserve	tuned no-feature baseline	yes	incorrect: attempts all 15 and consumes reserve
GitHub Octokit rate-limit handling	external	no	zero measurements; real client/response required
gh api rate-limit handling	external	no	zero measurements; real CLI/headers required
Envoy global rate limit	external	no	zero measurements; real service/configuration required

Repository inspection found no equivalent first-class fak integration beyond the native GitHub path, so no integration arm is fabricated. Mocks and package availability are not results.

Local native witness

go test ./internal/mutationbudget -bench BenchmarkEstimateMutationHour -benchmem -run '^$' -count=5

Windows/amd64, AMD Ryzen 9 9950X: 351.4, 361.4, 448.4, 431.7, 410.1 ns/op. Median: 410.1 ns/estimate, 368 B/op, 6 allocs/op. This is in-process policy cost, not GitHub network latency or API billing.

Reproduce

go test ./internal/mutationbudget -run TestCompareLocalKeepsAPIBudgetAlternativesExplicit
go test ./internal/mutationbudget -bench BenchmarkEstimateMutationHour -benchmem -run '^$' -count=5
go test ./internal/nativebench