The key point is that POSIX does not actually define a real, physical epoch, and it explicitly avoids doing so. POSIX says every day is exactly 86400 seconds, leap seconds are ignored, the relationship between POSIX time and UTC is unspecified. In other words, POSIX defines how the "time()" API behaves, not what the underlying timeline is. It’s an OS level contract, not a timekeeping standard.

The problem I’m addressing is that POSIX retroactively assumes a form of UTC that didn’t exist in 1970–1971 and it provides no authoritative mapping between POSIX time and actual UTC. That’s why you can’t use POSIX as a normative reference for "seconds since 1970". It’s not a definition of time, it’s a behavioural model for system clocks.

This document is trying to define the actual timeline that people implicitly assume when they say "Unix timestamp", in a way that’s historically honest and citable. When I was looking at a number of RFCs that use these timestamps they didn't cite this reference, but went with the wooly "Number of seconds since 1970".