Changes introduced at earlier minor upgrades can sometimes not kick in until after a subsequent restart, sometimes triggered by eg. a minor hardware fault sometime later. Had such a problem elsewhere only a week before and a total (power/off ) reload of the system didn't clear it (would only have precipitated it in fact - if it hadn't already shown up by then) . Realising your last resort has failed is scary.
Only a backup/redundant system with total autonomy from the main protected system can necessarily avoid running into this mess. Several mission critical operations have them. Can only wonder why this one didn't, or did they realise what redundancy means ?