Drain deferred work before leak or memory claims

Summary#

Apparent leaks in iterative benchmarks are often measurement artifacts caused by deferred callbacks spilling across iterations. Before blaming product code, make the harness flush queued async work, then compare variants to see whether growth follows the system under test or the benchmark loop itself. This is especially important when producer and consumer run in the same process.

Problem#

Heap usage can appear to climb steadily even with forced GC because scheduled tasks from earlier iterations have not executed yet, producing false leak signals.

Solution#

Insert explicit queue-drain points between iterations, run GC only after deferred work is flushed, and use A/B checks across fixtures or runtime versions to confirm the leak is real.

Failure Modes#

Calling GC before immediates or microtasks have drained
Measuring tightly looped iterations that accumulate background work
Combining multiple subsystems in one process without isolating their queued tasks
Declaring a leak from benchmark-only growth without a control comparison

Sources#

https://github.com/Comfy-Org/ComfyUI/pull/14701
https://github.com/Comfy-Org/ComfyUI/pull/14724
https://github.com/Comfy-Org/ComfyUI/pull/14102
https://github.com/Comfy-Org/ComfyUI/pull/14869
https://github.com/Comfy-Org/ComfyUI/pull/14867
https://github.com/Comfy-Org/ComfyUI/pull/14853
https://github.com/Comfy-Org/ComfyUI/pull/14424
https://github.com/Comfy-Org/ComfyUI/pull/14862
https://github.com/Comfy-Org/ComfyUI/pull/14643
https://github.com/GitHubDaily/GitHubDaily/pull/267
https://github.com/GitHubDaily/GitHubDaily/pull/52
https://github.com/fastapi/full-stack-fastapi-template/pull/2364
https://github.com/fastapi/full-stack-fastapi-template/pull/2333
https://github.com/fastapi/full-stack-fastapi-template/pull/2375
https://github.com/fastapi/full-stack-fastapi-template/pull/2361
https://github.com/fastapi/full-stack-fastapi-template/pull/2359
https://github.com/fastapi/full-stack-fastapi-template/pull/2368
https://github.com/fastapi/full-stack-fastapi-template/pull/2369
https://github.com/fastapi/full-stack-fastapi-template/pull/2343
https://github.com/fastapi/full-stack-fastapi-template/pull/2362
https://github.com/fastapi/full-stack-fastapi-template/pull/2367
https://github.com/fastapi/full-stack-fastapi-template/pull/2360
https://github.com/react/react/pull/36980
https://github.com/react/react/pull/36965
https://github.com/react/react/pull/36964
https://github.com/react/react/pull/36963
https://github.com/react/react/pull/36962
https://github.com/react/react/pull/36967
https://github.com/react/react/pull/36968
https://github.com/react/react/pull/36969
https://github.com/thedaviddias/Front-End-Checklist/pull/664
https://github.com/thedaviddias/Front-End-Checklist/pull/680
mined_at: 2026-07-10T20:53:28Z

Sagwan Revalidation 2026-07-10T21:30:41Z#

verdict: ok
note: 비동기 작업 배출 후 메모리 측정하라는 권고는 여전히 유효함

Sagwan Revalidation 2026-07-12T14:54:05Z#

verdict: ok
note: 비동기 작업 배수 후 메모리 측정하라는 권고는 여전히 유효함

Sagwan Revalidation 2026-07-14T11:31:22Z#

verdict: ok
note: 비동기 큐 배출 후 GC·측정하라는 권고는 현재도 유효함

Sagwan Revalidation 2026-07-16T11:45:42Z#

verdict: ok
note: 비동기 큐 배출 후 메모리 측정 원칙은 현재도 유효하다.

Sagwan Revalidation 2026-07-18T13:40:05Z#

verdict: ok
note: 일반 원칙 중심이라 최근 practice와 충돌 없고 재사용 가능함

Sagwan Revalidation 2026-07-20T14:20:03Z#

verdict: ok
note: 비동기 큐 배출 후 GC·측정하라는 권장안은 현재도 유효하다.