Telemetry: https://mzl.la/2rcR8Nd

Possible regression range from May 05 to 06: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=0b255199db9d6a6f189b89b7906f99155bde3726&tochange=37a5b7f6f101df2eb292b1b6baaf1540c9920e20

It would be nice to find someone to take care of this who isn't me. :-)

This regression is even more visible in the INPUT_EVENT_RESPONSE_COALESCED_MS probe: https://mzl.la/2qxhjx6, from 32.85ms to 38.14ms.

The INPUT_EVENT_QUEUED_APZ_MOUSE_MOVE_MS and INPUT_EVENT_QUEUED_APZ_WHEEL_MS probe also shows clear regression around 4/5 and 5/6 so I think we can narrow down the regression to APZ events. Telemetry: https://mzl.la/2q89QWA https://mzl.la/2q8rPMI

(In reply to Kan-Ru Chen [:kanru] (UTC+8) from comment #4) > The INPUT_EVENT_QUEUED_APZ_MOUSE_MOVE_MS and INPUT_EVENT_QUEUED_APZ_WHEEL_MS > probe also shows clear regression around 4/5 and 5/6 so I think we can > narrow down the regression to APZ events. > > Telemetry: https://mzl.la/2q89QWA https://mzl.la/2q8rPMI Actually that only means the events are in the queue longer since wheel and mousemoe events are fired frequently.

Backed out bug 1352889 today to see whether it is the cause.

[ni=kanru to investigate whether comment 6 helped]

The backout of bug 1352889 didn't make any difference. I'm going to reland it, but that means you need to look for a different culprit here.

See bug 1362094 comment 25. We decided this bug is INCOMPLETE due to the limitation of the probes.