All this boils down to is the ancient notion that system up-time is limited, and that applications must be quit by a "higher intelligence" (human intervention) in order to make room for other applications.
Today's devices and OSes are built for near-100% up-time. It's expected that, like a living organism, the OS will run 24/7. Otherwise, how can the system wake for an incoming phone call, receive an incoming text, or trigger an alarm? "Sleep" time is applied to backups, managing on-device storage/shuffling surplus data to the cloud, Machine Learning/Artificial Intelligence tasks that lead to accurately predicted app/data "suggestions" and scarily-good selections of photos for machine-curated albums. Geez, my Watch knows that at around lunchtime on workdays I'll want to run a 58-minute timer.
Today, the difference between quitting an app and leaving it running mostly comes down to "saved state" - that saved state is not stored in RAM, it's shuffled off to fast, Flash storage along with the rest of the app's code and data. While Flash is still much slower than RAM, it's fast enough (especially when on an SoC) that RAM requirements are substantially reduced (hence the A-series chips' relatively small amount of RAM) - less code and data needs to remain RAM-resident. A force-quit of an app simply flushes the saved state and brings things back to square one. That can be useful when the current state is unstable, but so long as the current state is good, no need to flush (it's not like "saved state" has an objectionable odor).
I remember the bad old days of MS-DOS, Classic Mac OS, and Windows pre-95. The OSes and systems were unstable, hence prone to crash. We were lucky to get a full 8 hours of stable uptime. A full, nightly shut-down was mandatory to ensure a clean slate in the morning (of course, we had to remember to Save before shutdown). There were insufficient system resources to constantly auto-save, so when there was a crash or a powerline hiccup, much could be lost. The notion of automatically re-opening all apps and documents after either a crash or an orderly shutdown was inconceivable.
But that changed when Unix-like OSes moved from mainframes and engineering workstations to garden-variety PCs, and inexpensive systems became sufficiently powerful and sophisticated that they could deliver weeks (or more) of reliable up-time, with enough system overhead to allow for continual auto-save/saved state. We began to take for granted that a system could crash but we could still resume work right where we left off.
Up-time is so good that the monthly free software update may be the only time a system is restarted. Those who prefer to avoid updates undoubtedly have to restart anyway. It's worth wondering what percentage of those who complain about the poor performance of their old phones have simply gone a few months without a restart.
OSes keep getting better and better at maintaining uptime; better and better at recognizing unstable states and isolating (and quitting) the rogue processes underlying the instability. In the foreseeable future even major software updates won't require a restart. And while Apple TV's Foundation bears little resemblance to Asimov's (and Apple TV's version introduces a Positronic intelligence much, much earlier in the narrative), we are getting closer to the day when an artificial "mind" might outlive a human's by thousands of years, or more, without a single reboot.