For me there are many real world cases which can't be handled without showing the user a fatal error.

E.g. - the developer asks for an runtime created asset that doesn't exist anymore or there's a condition under which a block of code runs forever (actually in this case the flash runtime stops after 15 seconds of activity and allows the user to kill it but it's still a "crash" insofar as the user is concerned.

But Flash often crashes way worse than that; it segfaults, or raises a structured exception. It used to take browser processes with it, before Chrome made it popular to have plugins in their own process.

I suspect Flash's (lack of) stability was one of the reasons for considering this design decision in Chrome.