Other extra-level logging methods I could make use of are: log.random (with argument(s) defining how random), log.context (everything the runtime knows at that point), and log.assert. These may be asking too much of a logging module, but on the other hand, if they lead to software that's easier to maintain, maybe they're worth bargaining for?
Anyway, I would love to hear your thoughts on how/why we ended up in a logging straitjacket of sorts. And yeah, I've defined ALWAYS as custom level before, and gotten it to work; I'm curious to know why it's not built in.
ALWAYS might not always be true. Can the message be logged if the hard drive and network failed?