I don't believe they are inevitable.
In fact, quite the opposite.
My personal email/FTP/Usenet/Shell account system has been online, and available to whoever has an account, since Flag Day. That was January 1st, 1983, when we changed from NCP to TCP/IP. It had already been running for a number of years, and probably would have survived the change, but I chose to reboot everything at midnight, just to come up from scratch.
Note I said "system" ... it's multi-homed, multi-OS, multi-hardware, multi MTA (and etc). ... redundancy is fitted in everywhere I can fit it. It started as a Thesis platform when I was at Uni (three locations: at SAIL, under Bryant Street in Palo Alto, and at MAEWest), and now is spread out on six continents.
Over-kill for a home system? Absolutely. But as a research platform, she's mostly tax deductible. It scales well, and parts of the concept are in place at several Fortune 500s. They should see similar uptimes for the bits they use, barring the almost inevitable catastrophic human maliciousness ... and even then, systems are in place to minimize that kind of damage. Maintenance at this stage of the game is on the order of minutes per month, and that's mostly just scanning the logs for anomalies.