I agree with you on the need for good dedicated test engineers, but even an apparently low crash rate can be damaging at scale. If you have 10,000,000 customers, a crash rate of 0.17% (over some unspecified time period) is 17,000 crashes (in the same time period). On the other hand, if the crash is deterministic, but caused by some user-specific state that affects 0.17% of your customers, that's 17,000 unhappy customers who are drastically more likely to post reviews than the potentially millions of happy customers not experiencing crashes (and perhaps reviewed the app positively years ago).Note: I don't use Facebook's app often, and have no context for what the actual issues are.
coldcode|11 years ago