It looks like what you were describing as compensation problems could actually be capacitive decay. If you say the trace looked o.k. with 330hz and a x10 probe but the x1 gave this bad display, I wonder if the x10 probe would look similar at 3hz input, because it would have a longer time to decay that you won't see at 330hz.
Try putting a capacitor in series with the x1 probe on CH2-4 and see if you can simulate the same condition. You probably will need a fairly small cap to get the same waveform. If this can be verified it means there is a cap in series with CH1 but not CH2-4 the way you have it now set. If the 4 channels are set exactly the same but CH1 give a different display, it may be a dirty switch/relay that shorts out the cap to go from for A.C. to D.C. coupling.