2. One doesn't have to fully get rid of systemd. It's "merely" a matter of breaking the dependencies on it, which is a rather small technical burden. A distribution could do that and barely notice (well, ok, other than the fact that it will now have to offer Unix'ish utilities it used to have and which systemd replaced; but those are optional too).
3. The more it expands to take over the entire lower part of userspace, the more it will squeeze out people who want to do work in that space. All those people are enough to sustain and grow a non-systemd ecosystem. I tend to assume developers will prefer it in the long run.
4. Longer-term: Non-monolithic dbus-utilizing lower-userspace facilities. These should be able to get to a state where one can switch from systemd to them with relative ease and without losing the useful capabilities of systemd. At least, that's according to my very rough understanding of things.
