Xlib is peculiar in that it doesn't really expose you to the protocol very well. For starters, it doesn't immediately write every request. Partly because it saves bandwidth on slow networks to not send lots of tiny requests that might end up in separate packets.

Partly because the Xlib API is pathological in making it hard for you to minimize requests yourself. E.g. a lot of calls to modify graphics contexts can be batched up, but that requires Xlib to not worth requests immediately.

So hence there being some value on testing with different flush behaviour.

But this is Xlib specific, not really X.