+Doug. I htink this is a first RMD bug with pictures attached, we can use it as tracking issue for the progress on RMD and quality of snapshots.

Great to identify such a URL. Presumably this would be wanting some timeout after the onLoad event (which has none currently). I can try playing with this.

I reproduced this with a build that does some logging in the SnapshotController. I received the first 3 log messages very quickly and then the content ready very quickly after them. The last message is for the MaybeStartSnapshot 7 seconds after the *Available* event. 

07-11 14:54:50.328 I/chromium( 2032): [INFO:snapshot_controller.cc(52)] XXXX   SnapshotController::DocumentAvailableInMainFrame - schedule delay: 7000
07-11 14:54:50.785 I/chromium( 2032): [INFO:snapshot_controller.cc(63)] XXXX   SnapshotController::DocumentOnLoadCompletedInMainFrame
07-11 14:54:50.788 I/chromium( 2032): [INFO:snapshot_controller.cc(71)] XXXX   SnapshotController::MaybeStartSnapshot
07-11 14:54:57.331 I/chromium( 2032): [INFO:snapshot_controller.cc(71)] XXXX   SnapshotController::MaybeStartSnapshot

Two thoughts here: 

1. I expect adding a 1 second delay after the *OnLoadCompleted* would have captured better page; but more importantly:
2. Why didn't the 2nd MaybeStartSnapshot capture better page? Does last N support two snapshots? If so, could there be some issue with overriding the first one?

[Update on 2 - expected that no snapshot taken after the one for OnLoad.]

Locally,I tried adding a 2 second delay after the OnLoad event and also enabled the 2nd snapshot for 7 secs after Available as a side effect. I do not see a good page offline even though both snapshot SavePage calls happen after I see expanded content on the screen. It looks like the SavePage calls are happening succesfully from these logs:

07-11 16:26:08.234 I/chromium(17265): [INFO:recent_tab_helper.cc(90)] XXXX   RecentTabHelper::DidFinishNavigation
07-11 16:26:08.273 I/chromium(17265): [INFO:snapshot_controller.cc(53)] XXXX   SnapshotController::DocumentAvailableInMainFrame - schedule delay: 7000
07-11 16:26:08.733 I/chromium(17265): [INFO:snapshot_controller.cc(64)] XXXX   SnapshotController::DocumentOnLoadCompletedInMainFrame - schedule delay: 2000
07-11 16:26:10.735 I/chromium(17265): [INFO:snapshot_controller.cc(78)] XXXX   SnapshotController::MaybeStartSnapshot
07-11 16:26:10.735 I/chromium(17265): [INFO:recent_tab_helper.cc(139)] XXXX   RecentTabHelper::StartSnapshot
07-11 16:26:10.763 I/chromium(17265): [INFO:recent_tab_helper.cc(169)] XXXX   RecentTabHelper::ContinueSnapshotAfterPurge - SavePage ...
07-11 16:26:10.895 I/chromium(17265): [INFO:recent_tab_helper.cc(179)] XXXX   RecentTabHelper::SavePageCallback  result:0  id:1518079707374636729
07-11 16:26:15.275 I/chromium(17265): [INFO:snapshot_controller.cc(78)] XXXX   SnapshotController::MaybeStartSnapshot
07-11 16:26:15.275 I/chromium(17265): [INFO:recent_tab_helper.cc(139)] XXXX   RecentTabHelper::StartSnapshot
07-11 16:26:15.313 I/chromium(17265): [INFO:recent_tab_helper.cc(169)] XXXX   RecentTabHelper::ContinueSnapshotAfterPurge - SavePage ...
07-11 16:26:15.432 I/chromium(17265): [INFO:recent_tab_helper.cc(179)] XXXX   RecentTabHelper::SavePageCallback  result:0  id:7659746194718148213

There may be some other issue here than snapshot moment detection. Maybe someone closer to Last-N should take a look at this one.

From encountering and exploring another bug (https://bugs.chromium.org/p/chromium/issues/detail?id=628029), I start to wonder if I was seeing a traditionally cached page in my experiments rather than an offline page. I repro-ed again on my N7 and I see the same spinner page rather than full page but I am not seeing the offline link indication in the omnibox so I think I just am not seeing what was saved as an offline page.

... the original bug detail shows the offline indication so don't think I am getting proper repro after all.

Another RMD/SnapshotController use case to keep in mind.

This bug is mostly about investigation, and might not make a good fixit candidate.  Removing hotlist=fixit

 Issue 671673  has been merged into this issue.

This is working as designed, but the design is not the best.
A better design which might fix this is to use the RMD resource percentage signal:  crbug.com/699313