Finite mine top-up coordinator proposal (target remains unchanged by this agent)

Apply fix.patch after /tmp/erri-mine-grant-1724/fix.patch and /tmp/erri-mine-transfer-1724/fix.patch. git apply --check passed against the current successor.
Register only new tests/test_infantry_mine_coordinator.py; five other tests are existing programs adapted to the expanded production interface. Eight focused programs pass via python /tmp/erri-mine-resupply-1724/run_tests.py; logs retained alongside patch.

Production changes: ERR_HQ, ERR_InfantryResupply, ERR_InfantrySupplyPool.
HQ tries existing infantry ammunition, then mounted ammunition, then one finite PMN4. Mine admission consumes only existing delivered <=120s complete mine reports with ReadyMines exactly 1; zero/unknown/multiple never establish demand. Original source/roster/radio/downlink/rendezvous/time/ownership checks remain shared. Optional original MineStock batch never delays normal magazine readiness. Its exact <=2 closed identities are copied once and paired with the SAME service m_Spent ledger, with unchanged 96-spent ceiling. No independent spent pool, new actor/vehicle, resource creation or engineer-attempt reset.

At physical rendezvous, select a permanently attributable original on-foot carrier with exactly one native-ready mine and a settled original startup grant (HQ.MineSupplyReady). Pin carrier across pre-submit retries. Leaf independently pins the exact existing carried mine and exact incoming original truck mine. At most one confirmed transfer ends the trip. An unsubmitted refusal may release its reservation; submitted identities remain spent permanently.

New service state7 retains both recipient LogisticsReservation and exact truck SupplyToken/held flag after failed/stopped unresolved mine submission. Owned waypoints and downlink are cleared immediately. PollTransfers releases matching claims only when leaf proves terminal custody or exact incoming deletion. Thus a second truck with a newer report after shared cooldown still cannot race the same carrier. Existing stopped/pool polling retains uncertainty without new orders. Pool diagnostics separately label magazines moved versus mines moved and include state7 among active claims.

No native run in this proposal; tests execute extracted production methods against native boundary doubles. Actual navigation, inventory capacity/transfer and multiplayer remain native qualification limits. Independent reviewer is testing removal of state7 as a negative control. No claim that this adds mine placement repeatability or mine resupply to zero-issue formations.
