This begs the question whether we could have a feature to disable
overcommit on a sysadmins', or the software's demand.
Doesn't it break a process that wants to preallocate _physical_ memory
on start-up, and reuse it, without allocating any further memory to
_guarantee_ it will not fail due to memory allocations over its lifetime?
If the OS is lying to such software that it has acquired physical memory, but
the OS will only try to allocate it when it _thinks_ the process needs it,
this process will still fail if the physycal memory is unavailable at
that time. Shame!