It looks as though the primary cause of the "bad" images may be that the cameras at that viewpoint are too close to your body. The closer the camera is to an object of person, the more the image breaks up in the way shown in the top corner of your lower image.
You can test this principle by holding a camera in your hand and observing the depth stream break up more and more as you move the camera towards your body.
You can reduce the camera's minimum distance (MinZ) by altering the Disparity Shift setting in the Viewer. As Disparity Shift is increased to reduce MinZ, the trade-off is that the maximum distance (MaxZ) reduces.
For example, according to Intel's camera tuning guide, a Disparity Shift of '0' gives a MinZ of 45 cm and an infinite MaxZ, whilst setting the shift to '50' gives a MinZ of 30 cm and a MaxZ of 110 cm.
It looks as though adjusting MinZ with the Disparity Shift may not have been the most suitable method in this case. Although it is enabling the camera to be closer to the body, the reduction in MaxZ means that the camera's forward view distance becomes too short and cannot see as far as the position where the body is.
The image of your tripod mounting setup shows that one camera is going to be closer to the body than the other one because its mounting point puts it further ahead. It looks as though the "bad" image is coming from the upper camera that is further back though, judging by the height of the cabinet and desk fan in the depth image. It is a strange case. The break-up of the stream is symptomatic of being too close to what is being observed, yet the camera is mounted further away than the "good" camera.
The worst of the break-up is in the background of the image. I would recommend eliminating the possibility that the upper camera is exceeding its MaxZ sensing distance by moving the tripod to the front of the top of the computer, so that the upper camera is in roughly the same position that the "good" camera is now. If the "bad" camera image improves in that position, that would add to the possibility that it is a distance related problem. D435 has greater depth noise over distance than the D415 model does.
There does not seem to be any camera related information in the video.
RealSense cameras cannot be sent for repair. You can file a claim for a refund or replacement. Before you do so though, I recommend trying a 'gold reset' of the camera to factory-new status. You can do this by downloading the 400 Series calibration tool - file name CalibrationAPI-126.96.36.199-Setup.exe - from the link below and then using this instruction:
This message was posted on behalf of Intel Corporation
Please make sure the cameras have the latest firmware and upgrade RealSense Viewer and the SDK to the latest versions.
You should also set the preset for both cameras to High Density.
Let me know how it goes.