It only affects channel 1. It is most likely to be provoked by overdriving and frequent changing of the measuring range (time and voltage). It disappears partly by itself, but not always. After it disappears, it can be provoked again.
Vibrations (knocking on the housing) do not change anything.
In this picture, both channel using the same signal source. (calibration output for Probe)
When I switch to 40kpt memory depth, it disappears immediately.
You channel 1 it's completely unreliable, this makes your channel 1 unusable, it makes sense that the signal it's present only at higher memory depths, more memory it's essentially more data/resolution.
Try to see the signal with no probes connected, just to discard a defective probe, but if is still there then you should use your guarantee and return it
But if you got it from China too, I'll probably try to fix it my self...
So you can check:
1. if the BNC it's properly soldered to the PCB.
2. if the shield can it's properly installed.
if the previous seems fine I'll try next:
1. Remove the shield can on your channel 1 and look for bad solder joins.
2. look for components with physical damage, ceramic capacitors with cracks or something that didn't seem right, you can compare it with your channel 2, it's all symmetrical and should be all the same.
3. look at the components around one of the big ICs with scratched top, marked as u38 (maybe u36).