I spent pretty much all evening continuing to look at why neither variant of the Terratec Cinergy T XS works. It seems that across the v4l-dvb codebase there is a significant inconsistency in what module is responsible for the i2c gate control. There are cases where the tuner driver opens/closes the gate. There are cases where the demod open/closes the gate. There are cases where the demod *only* closes the gate, presumably because the tuner or the PLL opened the gate.
I suspect what might be the problem is *nobody* is opening the gate, so the write commands being sent to the tuner are being dropped on the floor. In the case of both the mt353 and the zl10353, they both close the gate only, and the xc3028 driver doesn’t touch the gate at all. Combined with tonight’s discovery that the current em28xx driver detects i2c read failures but silently fails to detect i2c write failures, and that might be the root of the issue.
I also resumed work on the Evga InDtube, now that a user has kindly volunteered to do testing. I had done most of the original work back in January but couldn’t test the analog changes since my device was physically damaged (I discovered that it was damaged before it was even mailed to me).