HP-UX did not. Exec8 (OS1100) did not. What ever it was we ran on the 1401s and 360/30s (and 9300s) did not.
We manually zeroed core on the 707xs but even then we knew it was a wasted 3 minutes because that was only done before the firs run of the day and might not happen again for several days (because each daily cycle took several days in some offices).
MS-DOS and Windows (even still?) were notorious for not hurting "deleted" files.
Is the heartbleed bug not proof positive that it is not being done today?
On the contrary. Heartbleed is "proof" that memory IS cleared before being
assigned to a *process*. The data available via the vulnerability is
limited to data from the process itself, not from any other process on the
system. ie, Heartbleed can give up your SSL keys, but not your /etc/shadow
file.
If memory wasn't cleared before being allocated to a process, every
multi-user systems would be vulnerable to Heartbleed-style vulnerability -
just allocate some memory, and go reading. Eventually you'd get something
containing /etc/shadow or other data you shouldn't be seeing.
Within a process (ie, memory being re-allocated to the same process) there
are ways to achieve the same thing, however as there's generally no
security reasons for doing so, and as there is a non-trivial overhead, it's
not done by default.
Scott