A specific Spirit Island query kills the Space, and may sometimes crash the entire cluster

(Bug, Investigate , Priority: Critical, Test Status: No automated tests yet , Reported By Bad Link: Thing 3y284oe not found, )

System

Summary: Eric has a page that is doing a QL query that looks reasonable but is doing something pathological, with catastrophic effects. What's going on there?

The first time Eric ran this, it apparently killed the entire cluster: things got unstable until I rebooted everything. The second time, everything eventually came back to life, but it took a while.

The query in question was a simple "For each of 50+ Things, find the # of Things of a particular Model that were created by the same user":

[[_creator -> +$creator
  Feedback -> _instances -> _filter(_equals(_creator, $creator)) -> _count]]

Obviously, that is better handled now with withValueIn(), but still -- a query shouldn't be able to cause this sort of havoc. What's going on? What's the best lever, architecturally, to make sure that we can't get into excessively tight loops? (Which I assume is ultimately what's happening here.)