Ground loop is very likely and it may not even be ohmic -
could simply be something like large capacitance between
domains, a pumping effect to charge it, and then attaching
the clip creates an ESD-like (EOS) event.
Depending on the various ground paths you can sometimes
see hundred-volt sine wave on high-Z nodes, simply
apacitive coupling / division. A MOSFET gate would look
high-Z until this voltage ruptures the dielectric. Just as
an example.
Getting an isolation transformer to feed the 'scope and
earthing all of the measurement gear to the prime
reference point (take your pick) might be the way to
go. Also avoid hot-attaching stuff that has any current
carrying capacity (like 'scope grounds - tip with 10Mohm
and few pF, not so much).
You might like to survey ground, power, control test
points with an "independent" 'scope and see what is
present as you try to revise your ground scheme. It
may show you high level AC where you expect low
voltage DC / pulse signals etc. You would like to get
to a setup where every signal / source can be shown
to match the expected conditions. Proving it, entails
getting them all right along the way. Cut-and-try,
the "try" had better include observation of the "side
doors" with some pretty comprehensive coverage
since you have not yet found the entry point or the
mechanism, let alone established that there is only
one.