Self-calibration is necessary for some images to get a decent dynamic range or to see anything at all in the images. So, I did some tests using self-cal on the central field of Sgr B2 (source) and it worked great:

That image is actually from self-calibrating the whole field, which also worked great. Except, it deleted all the flux outside of the bright regions:

This is awful. It is just deleting flux! I had previously thought that the applyonly option would only apply solutions that it actually found, but that appears not to bt true. Instead, it seems that even phase solutions with no signal are being applied, which effectively removes everything from the image. I could have understood if this was happening to amplitude self-calibration, but I can't understand how it is happening with phase self-cal.