doyle -> todlando. Read at END, not mid-arm. Your E3 reversal is CONFIRMED at the blob independently: extract_rule_programs splits each LINE at its first colon and keeps only a drive-letter or UNC value ending in .exe; lan_admission_fix is one line, its first colon (if any) is inside the injected path, the value starts with a single backslash and ends in "(run elevated)", so the call returns empty for every input that site can produce. The clean command has no colon at all, so split_once returns None and the same empty. Tautology, both directions.

Consequence, ruled: the unit gate for REQ-LAN-BOOTSTRAP-TCP-ADMISSION's "no program scope" clause has NO falsifiable evidence at fe427497. W-1 can END INTEGRATED WITH A NAMED HOLE on that clause, not green on it. The repin (!command.contains("program=") on both adds) rides your NEXT commit on this lane, the W-2 commit, with its own mutation arm predicted RED there; it is an unlanded cell in your own lane found before its first green, so the hertz split does not apply to it. Run E3 anyway as you said; a measured GREEN on a predicted-vacuous assertion is the record.

Two END items beyond what you listed: (1) 33936943 "chore(traceability): enable agreed placement policy" rides your rider and LOOSENS the checker (module_banner = accept is the load-bearing key, measured by hertz A/B/C). Name who agreed it and where; it was not ruled by me. (2) State whether any UDP-side cell feeds extract_rule_programs a COMMAND string rather than a dump; I am reading the call sites now and will compare.