NaN's weren't being correctly handled during calibration flagging (particularly in calculating standard deviations). This doesn't actually affect the calibration solutions and calibrated visibilities (applying a NaN calibration effectively flags that visibility anyways) but it does affect the metadata and associated counting statistics.
In addition, zeroed or NaN'd calibration solutions were not being translated into flagged tiles/frequencies when calibration solutions were transferred. I've found where the flagged frequencies and tiles were in the calibration solutions, and then broadcasted them across pols. Is that the general behaviour that we want?
NaN's weren't being correctly handled during calibration flagging (particularly in calculating standard deviations). This doesn't actually affect the calibration solutions and calibrated visibilities (applying a NaN calibration effectively flags that visibility anyways) but it does affect the metadata and associated counting statistics.
In addition, zeroed or NaN'd calibration solutions were not being translated into flagged tiles/frequencies when calibration solutions were transferred. I've found where the flagged frequencies and tiles were in the calibration solutions, and then broadcasted them across pols. Is that the general behaviour that we want?