[aidenybai]

Why

A child may expose live state through a getter that the parent consumes in a template result. Treating that read as a bare state notification produces a false positive from no-pass-live-state-to-parent.

What changed

• Recognize the consumed getter result while retaining warnings for bare notifications, opaque or mutable dispatch, and unproven React providers.
• Integrate the current stacked base 1f7953c59095b1a43bbc0b0ca2a5854ac0eda690 through a normal merge, including its shared provider-proof and numeric-data performance repairs.
• Keep the finite provider and mutation guards covered by regression tests. Existing unsupported default React.useCallback dispatch shapes remain outside this exemption.

Validation

Head: 898e6d3c6496f94a83b22f0956ce40419a950197.

All local gates passed: prepare, full uncached test run, lint, uncached typecheck, format, generated registry, fresh five-package build, JSON report smoke, and target-rule strict fuzzing (500 configured iterations, seed 42, required target firing). Normal commit hooks passed; committed source and all 175 runtime files match the validated tree.

The shared cache has direct regression tests for query modes, call shape, initializer ordering and fresh analyses. Disabling either cache layer fails its work-growth regression test.

Remote parity was not run because the required Daytona credentials are unavailable; a local RDE checkout is also unavailable.

Local evidence: tmp/fpfn-regression-profile-2026-10-20/continue-20261022/stack-drift-review-v1/pr-1994-validation-v2/proof.json and pr-1994-normal-merge-commit-v1/proof.json in that directory.

Completed performance and regression review

Exact target and fixed general full-engine comparisons passed. Each workload includes activation plus five cold and five hot timing samples per revision and dedicated CPU captures; raw gzip artifacts and timing rows were independently read back. All four wall cohorts are stable under the fixed 10% plus 250 ms budget; total CPU and RSS manual flags are clear. Every complete unrelated diagnostic and all five project-context fields match. The general workload retains all 256 diagnostics. These are finite local measurements.

Workload	Cold median, base → head	Hot median, base → head
Authored target	464.8 → 463.7 ms	311.0 → 310.6 ms
General 128	717.8 → 708.7 ms	381.4 → 383.1 ms

Target-specific evidence: the current-parent integration preserves every complete diagnostic from the externally updated getter head.

All available exact-head checks are green, including the natural react-doctor run started after marking this PR ready. The live stacked base and main were rechecked. This PR is ready for review; merge the stacked parent first.

CPU sampling limit

One retained dedicated CPU profile contains one signed −1 µs sampling delta. CPU attribution is incomplete at those exact samples; the original raw profiles remain unchanged. Separate measured wall time, process CPU and RSS rows were independently recomputed and stay within the declared finite budgets. This qualification does not extend to other malformed samples or universal performance guarantees.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/eslint-plugin-react-doctor@1994

npm i https://pkg.pr.new/oxlint-plugin-react-doctor@1994

npm i https://pkg.pr.new/react-doctor@1994

commit: 898e6d3