To leave the 6502 behind you execute a single instruction: XCE, eXchange Carry and Emulation. It swaps the carry flag with the emulation bit — so CLC : XCE (clear carry, then exchange) writes 0 into E and enters native 16-bit mode, and SEC : XCE returns to emulation. That is the whole gate between the two worlds: a bit-swap. And because it is a swap, running XCE twice restores the original state — it is its own inverse over the pair, the seal axis hiding inside a mode switch.
The demo swaps carry into emulation (entering native mode), then swaps again — and the machine is back where it started: live demo
“Switching a CPU’s word size needs a mode register and a reset.” — the 65816 does it with one bit-swap instruction. cited
The border between two processors crossed by trading one bit for another — and crossed back by trading again. self-inverse
On the canonical compiler, XCE takes (C=1,E=0) to (C=0,E=1) — entering emulation — and a second XCE restores the original pair: