When using JMH, why is it a bad practice to have multiple benchmarks running at the same time?

Question

I have a benchmark suite written using JMH, and I would like to run these benchmarks in parallel. I have a sufficiently powerful machine with plenty of CPU cores, so I figured I would simply run multiple JMH processes and isolate each of them on their own core.

However, JMH actually enforces a single-process policy by using a lock file. I can work around it by setting a property -Djmh.ignoreLock=true, at which point I get the following warning:

# WARNING: JMH lock is ignored by user request, make sure no other JMH instances are running

Therefore my question - if I have $N JMH processes running on a CPU with $N+2 cores, and I have pinned each of the JMH processes to its own core, which exact risks am I exposing myself to? How will this affect the reliability of my benchmarks?

EDIT: The tasks being measured are 100 % CPU bound, there is no I/O whatsoever. I run with SerialGC to optimize for throughput, but also to help minimize the number of threads the JVM will require.