So many libraries depends on TS and a PR like that would break a lot of them. I would say they need to put it under a flag.

So... a PR with a flag? :)

* * *

In theory, yes.

In practice, here are a couple of examples:

A thread in relation to "replace `fgrep` with `grep -F`": https://news.ycombinator.com/item?id=33189503

How people feel about Python's handling of Python 2 vs Python 3: https://news.ycombinator.com/item?id=34227760

Semantic versioning isn’t really usable for typed languages as the majority of fixes would be considered breaking changes.

Being able to declare major changes in a version number says nothing about whether it's a good idea to make them.

they made the change from any to unknown in the catch handler so its not without precedent

The Boolean one at least is pretty longstanding: https://github.com/microsoft/TypeScript/issues/16655

I imagine there are a lot of edge cases to wade through before mainlining.