Currently memory written by writeProcessMemory() is not followed by a
memory barrier, which is clearly documented in writeProcessMemory.
However the consequence is that a thread on a different core may see
inconsistent memory.  It would simplify/optimize the debugger if at
least resumeSuspended() or resumeSuspendedThread() issued an acquire
fence and writeProcessMemory() issued a release fence to ensure that
resumed threads saw consistent memory.

Currently there is no known way of ensuring consistent memory for other
threads without introducing a memory barrier in the thread context
switch code.