About 10 years ago when they introduced the EX40 (I think?) those hard-froze randomly every couple hours to days on Linux. But only for some users. They couldn't track down the issue the first few weeks, I guess that's what you get for being an early adopter. They must have gotten (un)lucky during testing and only had setups that worked.
It was first suspected to be certain brands of RAM, so I requested a RAM-swap which unfortunately didn't help. Then a BIOS update which also didn't help. Then someone figured out that nohz=off on the KCL fixed the problem and I had it running like this successfully for a few years. Long after at least one dist-upgrade I remembered that and removed the option again, and the server still ran stable.
There's no real morale to this story I guess, but at least the support is super responsive, and as the root cause wasn't clear at that point didn't hesitate to swap random stuff if you requested so. Also had a faulty HDD last Sunday in one server and requested a swap, which they did within 20 minutes of me opening the ticket.