Let's see, let's see...
02:15 AM cron job that did not run one Sunday, some important scheduled payments thingy, with other things then running expecting it had run, oh lord the manhours cleaning up and correcting after that disaster. Yep, servers were in US/Pacific, nice little time bombs left laying around for months until the pointless time wobble sets them off...good luck testing for that, hence the modern "just put it in UTC" to eliminate that class of problems. Alas, many servers still get placed in stupid timezones, or being so cannot be fixed (risk too high, bitrot, etc).
Then, at another company, the on-call got woken up each and every DST wiggle, because, you know, credit card latency was now 3600.00049996 seconds, or something, and hey! send pages! at who knows what hour; can't fix that software, legacy stack you see. On a positive note, the next version did run everything in UTC, thank goodness, but not after a few years of pointless, stupid, dumb pages.
Finally, there was the 2005 or so "energy savings" act, when they changed when the time wobble happened, and then everyone was spelunking around all the codebases, finding out how many custom date-time libraries had crawled into the systems (hint: lots) that all then needed updating. Maybe they found 'em all? Who knows, some reports might still be off by an hour, sometimes. Worst was the Exchange system, good fun post-change as they eventually threw up their hands and said "here's your new, empty calendar, good luck" and then 10,000+ folks were scavenging through old email invites to try to figure out who had been going to what meeting that was no longer there.
Probably seemed a good idea at the time. Isn't.