Comment 4 for bug 1850709

A size of 64 megabytes shouldn't work, that's scarcely enough to run at all on x86-64.

Is it correct to say that with a 1GB heap, your failure is after a *very* *long* time?
You're not claiming a failure after like 10 GC cycles, right?

It looks to me that the 'pin' count is the problem. I'm able to produce a steady increase of pinned objects in generation 1 with that test loop, but so far it's been running for at least half an hour without crashing. I think it will crash though, I'm just wondering at what point.