As a good example, we're on the verge of ditching JQuery.
I always have to dive into the guts of a library to figure out bugs regardless - I feel like that's essential for a software engineer to become well versed in, regardless of whether the work is in frontend or backend.
The cost of not producing at a fast enough rate is something almost no business can afford, especially in the startup world. The speed at which companies have to move are always on full thrust because if they don't move fast enough, then it costs sales, reputation, and gives competitors more of an opportunity for market entry.
Homerolling your own framework/library can be a dangerous operation, and you lose the benefits of the lessons that others have learned in maintaining open source frameworks/libraries.
Yes. Only the code you wish wouldn't!
(seriously. I'm maintaining complex logic in ant. AUGH MY EYES)
1. Change of behaviour between versions, (e.g. the way a frameworks integrate with a page)
2. Reliance on functionality being supported into the future
3. Customer expectations
4. Browser support
That said, it is true that a frameworks gives you the ability to put together an app in a few weeks. However, when you know that what deal with with something that needs to be maintained for years than knowing what does each piece of your code is far more important.