From Harold's email, 1/21/2015:
This is the current list of "reasons" we have agreed upon to normalize
and do statistics of the flagging that is entered into PL flagging
templates:
outlier_amp: either low or high amplitudes (systematic difference in value)
outlier_phase: same for phases
noisy_amp: higher scatter in amplitudes (higher sigma)
noisy_phase: same for phases
delay: at least one wrap of phase v/s frequency in all spws
ant_position: same for the plot v/s time when field is a point source (over the observation time, although the timescale for this is not
really fully clear)
edges: noisy or otherwise bad amplitudes or phases at edges of spas
no_phacal: no phase calibrator after last science scan
quack: dip in amplitude at beginning or end of scan
tsys: high or otherwise bad Tsys
bad_rawdata: frequency combs, broad symmetric birdies or other artificial instrumental errors