Comment 6 for bug 1641241

SMT=off not fixing it makes me a bit suspicious of it being the same bug, but races are hard to reproduce reliably, so it still could be.

Of course, it could also be a double-unlock in your code, or Yet Another Bug elsewhere in lock elision logic, all of which would be masked by you having tested with my test packages above (which just disable hardware lock elision entirely).

As soon as we have a proposed patch for the Mongo-exposed bug, I'll be sure to post a pointer to test packages here, so you can burn those in and see if they resolve your issue but, for now, I'm going to continue treating them as two issues.