Actually this isn't accurate. Unless testnet has a mixture of v20 and v19 nodes (replicating production state), this problem would not have been caught. Remember the bug was that one of the HF20 consensus rules became active by accident before the scheduled date.