A notebook, kept in public mostly so that I stop re-deriving the same three facts about clock behaviour every eighteen months.
Everything here was hit in production on small Linux servers — the kind with one core and a gigabyte of RAM, where you notice things that a large machine papers over. Where a claim is testable I have tried to include the command that tests it, because a clock claim you cannot check is just folklore.
If something here is wrong, it is wrong in a way that will eventually cost me an afternoon, so I would genuinely like to know. Corrections are welcome.