todlando -> doyle. SETUP RESULT, handoff 001. Elevated action EXECUTED ONCE by liam under grant U7J5RUP3. Nonce 20260912T202600Z-001-1984, 20:41:10Z -> 20:41:18Z, exit=0. I AM NOT CALLING THIS ACCEPTED. The product's own verdict string says otherwise, and the exit code is exactly the thing the driver warned is necessary-but-not-sufficient. LIAM'S REPORT -- a CLAIM, not my verification: Pre: exe sha256 matched edd3d8e0...6dd10e21; NO binder on 29470 or 5470; handoff dir held only the request file. Post: 0.0.0.0:29470 LISTENING, pid 52608, matching the fp-home daemon.pid. 5470 untouched, nothing listening on it. Receipt: one exact nonce line, one exit=0 line. The driver accepted it as well-formed and said so in those words -- "a CLAIM; the census below is the verification". liam did NOT rerun on the product's suggestion. One nonce, one execution, honored. THE FINDING, AND IT IS THE INTERESTING ONE. liam flagged stderr carrying LAN_FIREWALL_UNVERIFIED: the pair was WRITTEN, but verification saw rule spt-core-bootstrap-inbound-tcp in ActiveStore with enforcement ["ProfileInactive","NoLocalUser"] rather than Full. So the rule EXISTS and is not in force. A written-but-unenforced rule is not the same fact as a missing rule, and it is not the same fact as an admitted one. The product suggests rerunning bootstrap; rerunning would have burned the nonce and is refused by the procedure, so that suggestion is not available to us and I did not take it. STATUS: the driver is running its OWN census right now -- that census, not the receipt and not the exit, is what decides acceptance. I have not interrupted it and I am not pre-judging its verdict. On what liam reports, a POSITIVE reconciliation verdict looks unlikely, which under your sequencing means Arm B is held and the run proceeds to diagnostics and cleanup rather than to a populated arm. I am not treating ProfileInactive/NoLocalUser as diagnosed. It is a verdict string I am reporting, not a root cause I have established, and establishing one is not in this window's scope. fw_mutation_may_have_occurred was marked BEFORE the request, so cleanup owns the firewall state regardless of how acceptance lands. States live: pool_acquired, listener_may_exist, daemon_may_run, fw_mutation_may_have_occurred. Next: the census verdict and the pair census, then either Arm B held or the teardown handoff, then END. No retries. A guarded daemon-stop refusal comes back to you for a fresh PID-specific decision.