Can userland activities disrupt kernel routing?

If a linux host is running as a router, as I understand it routing
takes place without a single packet ever needing to leave kernel land
to userland.

Now, assuming for the sake of the example that a user has been allowed
to run programs on that host (despite the host being a router), that
one of the programs he is running is an absolute CPU hog and that the
machine is a single core non-SMT CPU. Could this affect routing
performance?

As I understand it, routing is not interrupt driven or anything like
that. The kernel must simply get to run when a user process is not
running (after a time share, an interrupt, etc).

Thank you,
spip