When it meets the goals of the project. Tossing out legacy support probably means breaking behavior for people with the fewest resources. Larry Ellison probably has someone whose job includes making sure his laptop is up to date. A family living in a slum with a hand me down laptop doesn't.
Good luck.
I think you can throw out support for these legacy technologies IF:
1. Doing so results in competitive advantage--such as being able to move significantly faster, being able to maintain significantly cleaner codebase, etc.
2. Your library does something unique that no one else does - If there are tons of other libraries that do the same thing, you will be thrown into a feature war. I guess that's fine and there's nothing wrong with that but I personally wouldn't want to be in that position.