You're correct in that the only standards-based way to retain a POST in the session history is to not disturb an existing entry. However:
> it seems to try to say things are simple or easy
That's an unfair mis-characterisation. The developers are not pitching a universal panacea that solves all your problems and handles every edge case. They are offering an architecture that simplifies many common scenarios, and one that is thoroughly developer-friendly when it comes to supplying observability and integration hooks for edge cases.
For this latter purpose it merely remains to bother with reading the (clean & elegant) source code to enlighten oneself.
> it seems like it is undocumented
On the contrary, the behavior w.r.t full-page replacement on non-idempotent verbs is extensively discussed in the Turbolinks repo.
The "Turbo Drive" component appears to me as essentially unchanged behaviour in Turbo 7.0.0beta1 from Turbolinks versions 5.x. Turbolinks was introduced in 2013, has many years of pedigree and online discussion, and is well understood by a large developer community. Turbolinks was always maintained, even being ported to TypeScript (from the now venerable CoffeeScript) ca. two years ago with no change in behaviour. Turbo Drive is, practically, just a slightly refactored rebrand of the TypeScript port.
The stuff everyone is so excited about are Turbo Frames and Turbo Streams. These are new, and may be used without adopting Turbo Drive: as with practically everything from Basecamp, the toolkit is omakase with substitutions. They are, nevertheless, complementary, so you get all three steak knives in one box.