Context
The horizon quotients grow by surjective refinement. Once two adjacent global levels have equal finite size, the bond is bijective and no later observation can split a class.
Definitions
- counts residual behaviors on all future words of length at most , over all prefixes.
- A global plateau uses the complete residual orbit, not a finite prefix sample and not one fixed-length cut.
Hypotheses and scope
- The quotient counts are global over every prefix.
- The minimal-DFA conclusion imports the classical residual theorem recorded as FPRD-T22.
Proof or evidence
An equal-size adjacent bond between finite quotients is bijective. Its kernel becomes a finite-index right congruence, so the shifted letter actions descend and the quotient recognizes the language. Conversely, a finite residual orbit bounds all horizons. The square-length language separates small nonuniform cut width from uniform recognizability.
Verification notes
The M26 audit checks plateau persistence, same-level action descent, unique factorization, conditional finite-window reconstruction, and the fixed-slice counterexample. It explicitly rejects false finite-prefix plateaus.
Limitations
- A plateau observed only among prefixes up to a finite budget may split later.
- Uniform bounds on separate exact cuts do not bound the number of global residual states.
- Finite-window reconstruction is exact only when an independent residual-state bound is supplied.
Notes
The plateau is global over all prefixes. A finite-prefix plateau is not a regularity certificate. The square-length family supplies the direct fixed-slice versus uniform-recognition separation.