My reduction strives to drive home the core of an idea that extends both at higher levels (i.e. we have passive/active cycles as a society as well, the most simple example being workweek/weekend), and lower levels, like your bacteria example.
And yes, it's not about "one cell" vs. "all the cells", more generally it's about "higher-level systems" vs. "lower-level systems" where also some functionality serendipitously adapts/fits/evolves into one of those modes without specific regard about where it semantically sits in this hierarchy of complexity.
Regarding REM sleep, I think our methodology and terminology is a bit too crude, electrical "activity" means very little about the state of the brain. For ex. you have no access to most of your memories while in REM sleep, you can't even determine you're sleeping most of the time, and when you do, you usually forget in a minute or two, your cognitive function is that low.
Honestly the way you described REM sleep, it started sounding a bit like unit testing (inputs/outputs come from the repository, not from the user, and this process goes through the entire brain in order). :-) Which... actually might not be far from truth (in some aspects of it).