# Hahn T2 against a FRESH IBM record, kingston and marrakesh (rule written 22 Oct 2026 night, before any job)
Runs 1 and 2 ran under the same IBM records (kingston 18:45:38, marrakesh 18:21:43 EDT). Combined: kingston 0.922 [0.885, 0.959],
marrakesh 0.933 [0.892, 0.975], both BELOW, chi2/dof < 1; run 2 alone CONSISTENT on both, so no bill line by the repeat rule.
This run answers the vendor's remaining answer, "the stated T2 was hours old": fire only after IBM's record shows NEW T2 dates for
most qubits (median T2 age at the record stamp under 3 h), hahn.py unchanged, open instance of .ibm_quantum_key_3.
Rule: a chip's bill line stands only if this fresh-record run alone reads BELOW AND the three runs combine (chi2/dof < 2) BELOW.
Any other result: the chip is reported on the site with all runs side by side and does not go on the bill.
KNOWN-GOOD: median P(1) at tau = 0 at least 0.90 per chip. Synthetic test: synth_hahn.text (unchanged).