There were a few top crashes we noted as regressions in 5.0. They were all tracked.
We determined that these related to turning on PGO in March.
What we did learn was that this caused a regression in a few key signatures. We turned off PGO for b3 and discovered validated that we saw less of these signatures but an increase in a more generic signature EnterMethodJIT so we think the crashes just shifted.
We don't know if there is some net increase in EnterMethodJIT crashes as a result - note, this is the top #5 crash on 4.0.1. Addressing this type of crash is likely a many month initiative.
The crashkill team went through all the tracking firefox_5 crashes this morning and removed the tracking flag for many of them.
At this time there are no big top crashers we are concerned about.
Possible red flag - appears to be an increase in the number of plugin hangs but we need more data to validate this.
Christian pointed out that our beta period has historically higher crashes/ADU, which later comes down as the user base grows.
Based on that we look really good
We're currently (~1.5 crashes/ADU) much better than Firefox 4 was during the beta period (~3.2 crashes/ADU) and much lower than 3.6 and 4.0.1 final (~2 crashes/ADU)
It was pointed out that there are 4.0.1 crashes we know that we didn't fix that aren't showing up in the 5.0 crash data, some concern as to why we aren't seeing those and if they are lurking issues
It was mentioned it might be due to lower add-on compatibility, especially if more incompatible add-ons are binary
Beta feedback
Plug-ins (Facebook/PDFs/games/video/websites I suspect are also plugins)
Extensions
SLOOOOOW/Hang
Maybe something related to missing bookmarks but that's unclear.