the way we usually do it is to just use syscall[...]
I'd personally be happier to try implementing the higher precision timeouts with something like a timerfd, before considering using a syscall wrapper around `epoll_pwait2`.
Of course it could be done, just define all the relevant syscall numbers, convert to a 64bit timespec, and do the `ENOSYS` test. It just feels like a dubious use of time which also adds some code bloat/complexity.