The PCs' statblocks are not absolute. They change as the players change them--in the ways they are permitted to do so. Sometimes those changes are diegetic. Often, they are not. We simply handwave away the inconsistencies by admitting that the abstraction is imperfect.
As for the other point here, that it would be unacceptable to change the PCs' statblocks: Of course it would,
for gameplay reasons. There is nothing
diegetic about not changing the players' statblocks. There is, however, something extremely important in the gameplay, the "ludus" part of the experience, namely that that pulls the rug out from under the player so they can no longer make reasonably-informed, meaningful decisions.
Those statblocks are a necessary input. One of several, to determine
what relative thing we are considering. It's not the only one. The rules themselves also provide an input, as do other elements.
It may be non-negotiable for you, but it is not possible to achieve within the rules of D&D and the other requirements you've placed.
We have to accept at least one of the following:
- We redesign the game so that it isn't what D&D does anymore, so that perfect mechanical symmetry between PC and monster is possible while still having entirely diegetic processes.
- We break perfect mechanical symmetry, thus allowing the abstractions (read: NPC statblocks) to change as needed to reflect the new contexts that PCs find themselves in, through diegetic processes.
- We abandon the requirement of diegetic processes, so that things change simply because the mechanics say they must change within the space of perfect mechanical symmetry.
The first is induced by the fundamental problem of preserving the basal, down-to-the-metal framework of a wargame....where we have
changed it so that one side must succeed not just for this battle, but for
every battle into perpetuity in order for play to proceed meaningfully. In other words, we have broken the symmetry ourselves by making PCs that must win and win and win and win and win in order for play to proceed, but NPCs that
only need to participate once.
D&D, at its very heart of hearts, is
designed as an asymmetrical game. It has to be. If it weren't, functionally 100% of campaigns would fall to TPKs. Even if the players have a 95% chance of avoiding a TPK with every "Deadly" battle, characters need
at least five "Deadly" encounters in order to gain any given level in 5e, for example. That's roughly 50 battles just to reach level 11, which is (very VERY roughly) the point most campaigns end. That would mean only ~7.7% of all campaigns would actually reach level 11 without a TPK. More than half of campaigns would end by
level five. (The first two levels only need ~3 Deadly combats each, so about half of campaigns would collapse by about that point.) And yes, I know you're of the opinion that repeated TPKs shouldn't be a problem for groups,
but they demonstrably are and that isn't changing.
The other two points--breaking perfect mechanical symmetry, and abandoning diegetic processes--have their own share of issues, I agree. It would be
wonderful if we didn't have to make this choice. But from where I'm sitting, I can't see any other way out. D&D's design, plus the twin requirements of exclusively diegetic processes and perfect player-nonplayer mechanical symmetry,
force us into an unwinnable scenario.
We are left with a trilemma. Either we change what D&D is, which no one is willing to do; or we abandon diegesis, which thoroughly offends the simulation fans; or we abandon perfect mechanical symmetry. Something's gotta give--unless you can find a way to square that circle.
The truly funny thing is, the whole "D&D is simulationist" schtick only surfaced after 4e. It was an edition warrior dog whistle to try to exclude 4e from the D&D family. That has somehow morphed over the years into this idea that D&D was ever designed with simulation as a priority. It never, ever was. 
