Which is an inconceivable screwup if true.

A decade ago I've replaced tx processing systems 2 orders of magnitude smaller than this in terms of money flowing through them that were 20 years old (i.e. older than the ECB was at that time), most often clustered across many servers, everything triple redundant per site and have at least a fully sync'd hot site for fallback in case an entire datacenter gets EMP'd.

So "hardware failure" is either a lie (most likely), or the competence is nowhere to be found (still kinda likely.)