In mixed version 4.0/4.2 clusters, both mongos and shards will be triggering autosplits for chunks. In the case of this test, sometimes mongos will trigger a split for a chunk that no longer exists (because of concurrent splits or because its routing info is stale), but when it does this, it prevents the shard from acquiring the dist lock when it attempts to do the split for the chunk we're actually expecting to split. So, the shard gets a LockBusy error on trying to acquire the dist lock, and we don't end up with as many chunks as we were expecting. This behavior is not incorrect, but it is difficult to avoid in this test, so we can just blacklist it for the last stable suite.