When I was at Google, I had to chase around a major regression in only Firefox. Rendering of a part of the site worked great in Chrome, great in IE, great in Safari... Tanked to seconds-per-update on Firefox.

I was ultimately never able to find the root cause, and the issue went away with the next minor update of Firefox itself.

The thing is: for a junior engineer (who are often the ones who get assigned that kind of task because it's not high-priority to the product's feature list, which is the primary driver of success), that's a hell-project. Ultimately, it's a negative mark on performance review: "I spent a month chasing a problem with a browser used by 2% of our userbase and the end result is the problem fixed itself" is a terrible waste of company resources. And since Googlers get to choose what they think will be most impactful, it should be no surprise they avoid that choice every chance they get.

Google's whole self-management approach tends to optimize towards the center at the cost of the fringe.