Comment 50 for bug 1797581

Bumped to field-high as we ran into this again in testing.

We have a workaround, but it's to not use 2G VM's, which is really silly and hard to remember when we go and add new deployments, especially because the failure mode is not obvious at all.