yes, it looks at a fine-grained clock, usually a cycle counter provided by the CPU for this purpose, to aggregate total on-cpu time for each process.

I think a lot of modern kernels are “tickless” - they don’t use a timeslice timer, and only context switch on IO interrupt, process yield, or when timeouts are specifically requested (including capping cpu-bound processes). https://en.m.wikipedia.org/wiki/Tickless_kernel

Surely you mean 0.0999999931082% ?