Does the on-disk image of an executable ever change?

[Michael ODonnell]

Thanks to all for the feedback and sorry about not responding promptly
to questions.  I'm required to be a bit coy here but let's say that
this incident got me investigating stuff upstream from the systems in
question and it's now pretty clear that they were not compromised (in
a malicious way) and that some changes need to be made to some of the
distribution practices in place.

> There are other kinds of corruption beyond a bad disk, though.
> Logical corruption in the filesystem doesn't need to be due to a
> bad disk, for example.

Right, lots of ways to hose your data: eg. bad unmount hygiene, kernel
bugs, scribbling into the raw device, bit errors introduced during
unverified copy ops, etc...

> Do these various "corrupt" binaries actually seem to work?

The few that I've tried?  Yes, I guess.  Often hard to tell, of course.

> I'm wondering if, somehow, an RPM transaction didn't commit
> properly.  [...]  If you suspect that might be it, one way to
> "fix" it might be to reinstall every package on the system.

Experimenting with that right now.