Turning on --enable-viz moves a large/complex piece of code from the browser to GPU process. Any bugs in this code that cause browser crashes will instead cause GPU crashes. GPU crashes are recoverable and the user experience should be better. We should do some measurements to get an idea on the potential improvement here.

1. We can look at crash reports for the browser process that originate in specific source files that are only used in the display compositor. These crashes would presumably happen in the GPU process instead of the browser process. This gives us an upper bound on the number of crashes we make recoverable.
2. When we are confident enough that --enable-viz works for a finch trial (presumably on Windows or Chrome OS) we can compare browser crash rate directly. This will give us a much better idea of the potential improvement.