Batch 241 (docs) ACCEPTED -- the batch protocol revised (one registered `sorry` per batch, registry in HANDOVER.md) and ROADMAP restructured: panorama + live-item index, Part II kept

No Lean change in this batch; the tree is as of batch 240-A (`9bdecda`).

## The revised protocol (owner's standing instruction, 2026-09-22)

One batch does exactly one of:

* (a) it eliminates ONE registered `sorry`; or
* (b) it eliminates ONE registered `sorry` and creates ONE new `sorry` in a NEW file --- entered in the
  registry --- which the next batch attacks.

**`HANDOVER.md` is now the SINGLE REGISTRY of the registered `sorry`s** (§1.1, each with its exact
statement); the `README.md` `sorry` map leads there; `ROADMAP.md` INDEXES the item the campaign is
attacking and marks it resolved only when every `HANDOVER.md` `sorry` is discharged.  Recorded in
AGENTS.md (ground rules + the acceptance checklist), README.md (the per-batch policy paragraph) and
HANDOVER.md §0.

## ROADMAP.md

* §1 **the panorama**: Theorem 1.1 and its four assumed inputs, and the live campaign's internal map
  (Theorem 8.10.2 <-- (8.10.8) = 8.8.9 <-- the §8.6 engine; Theorem 8.6.1 <-- (8.4.25)) --- ONE engine,
  two currencies;
* §2 **the live item: FLM Theorem 8.10.2** --- status is an INDEX into `HANDOVER.md` (no duplicated
  detail: item, branch, status OPEN, the resolution rule, the acceptance gate);
* §3 **the batch protocol**;
* §4/§5 the reading guide and the working agreement (kept);
* **Part II (the branches) is KEPT**: Streams S/T/M/A, the long campaign and the FLM campaign, with the
  campaign's closed-phase table and its open entries R1--R5 updated --- R3/R4 marked removed from the
  ledger (batches 239-A/240-A), R1/R2 re-pointed to the registered `sorry`s, R5 to `HANDOVER.md` §6
  item 3, and the dead `§1 row N`/`§6 item N`/`frontier #N` pointers re-pointed;
* the removed front-matter status detail (the old `## 1. Project status` tables with their per-batch
  notes) lives in `HANDOVER.md`, `NOTES.md` and the git history (`git show 9bdecda:ROADMAP.md`).

## HANDOVER.md (the registry, refreshed)

* "State as of batch 241"; the `sorry` policy paragraph is the revised one above;
* §1.1 now states the TWO registered `sorry`s exactly --- `MM.FLM810T861S.crossContraction_evenCocycle`
  (the (8.4.25) cross contraction; by batch 223-B IS the window-synchronized (8.5.13)) and
  `MM.FLM88T889S.theorem889_faithful` (`FacesAtStates`) --- and records that the 8.10.2 file has no
  `sorry` of its own (the old #3/#4/#1 entries were obsolete);
* the delivery table's 8.10.2 row says "no hole of its own"; §1.1's rows 3/4 are compressed to their
  removal records; §6 is rewritten: TWO `sorry`s, both the SAME §8.6 engine, with the assembly as the
  next sub-target and the six boundary inputs as the last item.

## Also

* README.md: the documentation map's ROADMAP/HANDOVER rows describe the new roles;
* all `frontier` references to the removed tables are gone from the four docs.
