Paired bird surveys identify linear relative detection drift only when time varies within calibration cycles under free method intercepts

We test which paired bird surveys can separate linear detection drift from community change. With free human and recorder intercepts, relative drift is identified exactly when time has a nonzero contrast around a cycle in the pairing graph. A connected graph can fail this test. Free pair interactions require repeated pairs. Absolute drift still needs an external reference constraint. An audit of 130,303 standard BBS runs in 1966-2025 finds the same observer on the same route at 52.3% of eligible five-year endpoints, 34.4% of ten-year endpoints and 15.1% of twenty-year endpoints. A conditional CTI precision example needs 769 completed ten-year pairs for a 95% half-width of 0.005 C per decade when contrast SD is 0.05 C and the reference is known. Reference slope error adds a floor that more pairs cannot remove. These are design results and assumed-noise scenarios. We do not measure hearing drift, ecological lag or field calibration performance. Code, proof, source hashes, aggregate counts and checks are included.