Spreading services across several small boards - clever or just overcomplicating things?

I did the small board stack thing for about two years, four Pi 3s and a Pi 4 doing home automation, network filtering, file sync and a camera feed. What wore me down wasn't the philosophy of it, it was the small stuff. SD cards quietly dying every 8 or 10 months, one board that would drop off the network whenever the spare room got warm in summer, and a shelf of wall warts that all hummed at slightly diferent pitches. Moving it all onto a secondhand Dell micro PC with a 512GB SSD took a weekend and cut my power draw too, since the four boards plus thier supplies were pulling more than I expected at the wall. The honest downside is what your mate said. When the Dell's PSU went last spring, everything stopped togather, and I was without heating schedules for a day untill a replacement arrived. But I'd had similar outages on the stack anyway, just spread out and harder to diagnose, becuase each little failure only announced itself thru wierd symptoms elsewhere. The watcher problem is real though. Mine runs on the mini PC itself, which is a bit of a joke, the guard watching its own prison. A cheap old phone or tablet pinging the main box from a different power circuit is prolly the actual fix, and I keep meaning to set that up. If you enjoy the rack and the beach names, keep it, but don't fool yourself that it's more reliable per watt.

I went the opposite direction from you a while back, and I'd do it again. Had a similar shelf of boards doing seprate jobs, and consolidating onto one used Lenovo tiny box with Proxmox was the best home lab decision I've made. Each service lives in its own container, so one crashing doesn't take the rest down, and the whole thing pulls about 15 watts at the wall according to my plug-in meter, less than two of the boards plus their supplies ever did.

The part that sold me was backups. With the boards, imaging an SD card properly meant pulling it and sitting at a laptop. Now the whole setup gets snapshotted nightly to an old USB drive, and restoring a botched update is a five minute job.

On your watcher problem, that's real no matter which way you go. Mine just pings me through a service when things go quiet, and the check runs externally, not on the box it's watching. A dead watcher that can't tell you anything is a dead watcher either way. The honest tradeoff is what your mate said about single points of failure, but a cheap spare board or a known-good backup image covers most of it. Power supplies failing one at a time was never the comfort I thought it was, they always seemed to die in the same week anyway.

Still finding my feet here so not sure if this is a daft suggestion, but I sort of have a third option going. My setup runs on an old laptop that was gathering dust, and the battery turned out to be the accidental win. We had a power blip last month and the whole thing just carried on like nothing happened. Also handy that it has its own screen and keyboard when something goes wrong and I need to poke at it directly. Downsides are it looks untidy on the shelf and the fan kicks in at random moments. Is that a silly middle ground or do other people do similar?

I've only got the one little board so far, it records my doorbell camera stuff, and even that one confuses me when it plays up. The beach names thing sounds fun though, mine is just called 'pi' because I couldn't think of anything better. How do you remember which beach does which job when something goes wrong, is it written down somewhere or do you just know it off by heart?