Report of a panic in the tcp syncache timer code due to another thread
sleeping while holding an inpcb mutex. Probably the IP stack is
sleeping somewhere while holding the mutex (under high load), which
violates assertions regarding the use of mutexes. If this can be
reproduced using WITNESS, it should be easy to debug.