I just submitted a feedback report under this username.

By the way, the restart happened on 5/24/17 at ~11:00 pm ET.

Thanks for the feedback report! It's at http://feedback/#/Report/61637227067.

It's generally best to file the feedback report before shutting the system down, but we happen to include the previous power manager log as well. :-)

Chrome started reporting video activity at 22:48:31 last night, and an audio stream was opened a bit earlier at 22:47:29. It continued until the lid was closed at 07:09 today:

[0524/224729:INFO:activity_logger.cc(20)] Audio activity started
...
[0524/224831:INFO:activity_logger.cc(20)] Video activity reported
...
[0524/225131:INFO:activity_logger.cc(20)] Video activity ongoing; last reported 5 sec ago
...
[reports continue]
...
[0525/070031:INFO:activity_logger.cc(20)] Video activity ongoing; last reported 1 sec ago
...
[0525/070331:INFO:activity_logger.cc(20)] Video activity ongoing; last reported 1 sec ago
...
[0525/070529:INFO:activity_logger.cc(20)] Audio activity ongoing
...
[0525/070631:INFO:activity_logger.cc(20)] Video activity ongoing; last reported 1 sec ago
...
[0525/070829:INFO:activity_logger.cc(20)] Audio activity ongoing
...
[0525/070907:INFO:daemon.cc(514)] Lid closed
...
[0525/070907:INFO:state_controller.cc(912)] Ready to perform lid-closed action (suspend)

So this was working as intended from the power manager's perspective. Do you happen to remember what you had open last night? The video activity reports indicate that there was a region 333x250 or larger being updated at 15 FPS or higher.

Per further discussion, it sounds like this may have been triggered by a (300x250?) video ad that was left onscreen. While 300x250 shouldn't be large enough to trigger the video detector (since its threshold is 333x250), perhaps a larger ad could've caused this. Note that an offscreen video ad shouldn't be able to trigger it, as far as I'm aware.

If you can find a site that reproduces this issue (e.g. in your browser history), it'd be helpful for debugging. We don't log this data, but I think that chrome://media-internals/ may have details about streams while the problem is ongoing.

Regardless, I think it'd be reasonable to reevaluate the thresholds for the detector. I believe I chose the old values based on what YouTube and/or Vimeo were using at the time.

YouTube's resolution guidelines are at https://support.google.com/youtube/answer/6375112?hl=en. The smallest player size they mention there is 426x240, which matches what I see used in a narrow window.

The first Vimeo video I clicked on was 960x540. It appears to shrink down to at least 392x221 in a narrow window, but it's probably rare for users to be watching such a small video for an extended amount of time instead of using a bigger player.

Note also that <video> elements (as opposed to Flash-based players) will probably keep the screen on regardless of their size. Chrome's update-based video detection was added back in the days when all video was Flash, and we now have an independent mechanism for inhibiting power management when <video> or <audio> elements are active. I don't know how it interacts with <video>-based ads, but it's easy to check if you point me at some URLs that you suspect are keeping the system awake.

Dan going to assign your way to determine if there's anything we can/want to do to improve powerd's video detector or if its WAI.

Sure. I'd still like repro steps for this, but I think I should increase the minimum video dimensions in any case.

This just happened again Sunday evening. What would you like me to do when I see it in action. This actually happens pretty often. Maybe a few times a month without me evening paying attention to it. Thus, likely even more often.

Send a feedback report via Alt-Shift-i and paste all of the open URLs into the description field, please. If you can look at each active tab and see which one is displaying video, even better.

I'm closing this, but please comment or reopen it if you can supply the info from #7.