actually i can't answer your questions
in practice -- when a new system is designed -- in ideal all relevant parties (providers of different modules) have meetings where they agree in detail the physical signal levels, timing and software data interfaces, formats, timing -- test it on site -- fix bugs -- repeat the previous -- until they're convinced the system operates
if by debouncing you mean buffering or quing the signal for further processing -- then obviously it is "valid at the time it's fixed , buffered or set into processing que"
on the other hand in science the "raw value" denotes to value that is an immediate reading or measurement value that has not been affected but that still has it's errors -- in that sense if you fix, buffer, record, cue a digital value it is still a raw value - but if timing is relevant - it is not (raw) because it's a vector (value,time) where time (a relevant moment of it) is something you actually won't capture for further processing so *it's never raw at the very next moment (*the time) thus nor the vector (value,time)