This isn't blocking from my perspective. It's a nice-to-have.

I'll keep it marked as "Blocking" the eng tracking bug for now so we can keep track of it in our burndown list, but let's deprioritize this compared to the other blocking bugs.

This isn't strictly blocking, so I'm going to remove the "blocking" label. However, it would be great if we could land this and merge it back before 66 rollout. David, when you're back in office, let us know how complex of a fix this looks.

This is not actually related to Dice, but is a regression:
Same thing happens with Dice disabled on trunk, but does not happen on stable.

I found the cause:
1) the sync confirmation dialog interprets closing the browser the same way as pressing "Undo"
2) When "Undo" is pressed, the user is signed out.

The logic is the same on stable actually and was not changed in  along time, and so I think this is WAI.

There is a subtlety though:
* if Chrome is closed *cleanly* (which includes Ctrl-C apparently), we go through the "Undo" flow.
* If Chrome *crashes*, then the signout does not happen, and we see the authentication error I was expecting

I think this logic makes sense. I was not expecting Ctrl-C to be a clean shutdown, but apparently it is, and that's probably fine.

Lowering the priority to P-3, in case we want to change something, but we may actually want to close this.

David, do you know how this gets logged with our user actions/histograms? Do we log a "Signin_Undo_Signin" user action before the shutdown takes place? Or do we log "Signin_Abort_Signin"?

I'd just like to make sure we have a sense for how often this is occurring in the wild, if possible.