The expected noise level in the image can be computed from the radiometer equation given the system sensitivity and integrations time; unexpectedly high noise levels are an indication of problems

Noise will be roughly additive and present where the data is, and thus can exhibit the same sidelobe pattern as the true sky signal

Be very careful in self-calibration on weak sources using short solution intervals - fake sources can be "manufactured" from noise!

Note that CLEAN or other deconvolution algorithms will wreak havoc on the noise distribution, and care should be taken in interpreting noise in restored or model images

Always know the expected noise level in the image given the system temperatures, atmospheric conditions, and integrations time!

Astrometry requires special care in calibration, as closing phase errors and the self-calibration process itself will corrupt the source position