FLM 8.6.1, batch 433-A (END STATE (2): the ORPHAN IS GONE --- the corrected per-(profile, distribution) spelling is REFUTED at two data and the batch-426-A piece 1 is DELETED; the chain's registered content is now EXACTLY (P) and (T), and `Definitions/FLM861PerProfile.lean` is sorry-free; FinalCheck's deficit row updated in the same commit)

REFUTATION (compiled, sorry-free, three standard axioms): the (b') repair 432-A prescribed --- the iterate identification with the A-side at the ROW-SUM-REDUCED state --- is FALSE at two inhabited instances of its own domain: (i) the 432-A datum, NO HYPOTHESIS: `as=(b,-b)`, `r=(1,0)`, `bs=(c)`, `s=(0)`, `u=e00`, `v=0`, `tot=0` gives `redListA=[(b,-1),(-b,-1)]` = the VACUUM and `expShift = -1`, and at `(p,j)=(0,-1)` the LEFT side is `(Y861 (iotaVL c))(-1)`, whose value on the vacuum is `iota(c) != 0`, while the RIGHT side is `0` (the z_0-power is 1, `prodExpansion_vacuum_left`); (ii) the k = 1 datum with `R.form b b != 0`: `redListA = redListB = [(b,-2),(-b,-1)]`, `expShift = -1`, and at `(p,j)=(-2,-1)` the LEFT side is the landed `iteratePow_k1_p = (R.form b b)•1 != 0` while the RIGHT side is `0` (`prodExpansion_neg_one_p`). Compiled collapse `MM.FLM8.correctedPerProfilePiece_imp_perDist` carries the falsity to the per-(profile, distribution) spelling, so NO per-distribution piece can carry (P): the (8.6.15) box sum is irreducible. Witnesses in NEW `Refuted/FLM861PerProfileCorrectedRefuted.lean` (mandatory header + `**Lean witness:**` line; `not_correctedPerDist_c1_instance`, `not_correctedPerDistPiece_432A`, `not_correctedPerDist_k1_instance`, `not_correctedPerDistPiece_k1`, `not_correctedPerProfilePiece`); index row in `Refuted/README.md`. REMOVAL: the then-orphan registered piece `MM.FLM8.perProfileCarrier_iterate` (unconsumed since 432-A) is DELETED from `Definitions/FLM861PerProfile.lean`, whose header now quotes its statement and states what re-registration would require; `Definitions/FLM861CarrierPieces.lean` changed in PROSE ONLY ((P)/(T) byte-compared against `git show HEAD:`). THE TEST IS NOT THEORY: 433-A's own truth is NOT decided by the removal (recorded honestly: at the 432-A datum its iterate side is the open (8.6.6)/(8.6.9) content, `MM.FLM8.B433Decider` in `Probe/B433A_blocks.lean`); the deletion is the owner's standing no-orphan remedy. CHAIN REGISTERED CONTENT: exactly (P) and (T) --- `#print sorries MM.FLM8.theorem861` names those two before and after; the chain's own `declaration uses 'sorry'` warnings drop 3 -> 2. `FinalCheck.lean`'s deficit row rewritten here (it had listed the deleted piece): it now names (P)+(T), the 432-A/433-A refutations with their witnesses, and the campaign's seven refuted spellings. Acceptance reproduced by the dispatcher: `lake build Definitions.FLM861PerProfile Definitions.FLM861CarrierPieces FinalCheck` EXIT 0 (9228 jobs, 0 error lines); `lake env lean Refuted/FLM861PerProfileCorrectedRefuted.lean` EXIT 0 (0 errors); forbidden tokens: the refutation file AND `Definitions/FLM861PerProfile.lean` both sorry=0 admit=0 axiom=0 native_decide=0. NOTES.md section 394.
