The fix on
DM-16392 is easiest to confirm on the pipe_analysis processing outputs of the RC2 bi-weekly reprocessing. The issue was that the order aperture corrections were being added to the schema was not deterministic (loop was being done on a set). As a result, when attempting to concatenate all patches of a given tract into a single catalog and using a template schema, the aperture corrections were often pointing to different key offsets, thus getting the wrong column appended into catalog. This issue was discovered in perusal of the plots output by pipe_analysis's compareCoaddAnalysis.py and compareVisitAnalysis.py when comparing outputs from one reprocessing to the previous one (between which no algorithmic changes had been made that could've caused this discrepancy, backed up by the identical flux measurements for the same points). This ticket is to capture the effort that went into diagnosing the bug and ensuring the fix was successful after the w_2018_44 – assuming the fix lands on time – is completed.