We can consistently reproduce the hang using 2 specific mongodump bson files. When running the restore on an entire database, we noticed a hang in the same approximate location, so we switched to "-j 1" to find the specific file. We found at least 2 that can reproduce the problem every time.
Things we've already tested:
- Regenerated mongodump
- Used db.copyDatabase() to validate that the collections at issue copied in whole (they did)
- Performed the dump and restore to different members of the replicaset
- Attempted to perform the restore in windows vs unix, without success.