It's amazing how frequently this seems to happen. A lot of the huge downtime events I've read about have occurred because of failures in the systems used to rec

It's amazing how frequently this seems to happen. A lot of the huge downtime events I've read about have occurred because of failures in the systems used to recover from failures. It seems that perhaps more time/effort needs to be spent testing the systems (and the use of those systems) which are critical for handling failure. While building them, it's easy to dismiss their scaling requirements on the basis that they shouldn't ever be under significant load.