I would guess absolutely that the thing will ordinarily be receiving DMX; the PI/PO is only needed if one wants to change the address of the thing.
The pictures I found show PI/PO daisy-chaining, as I noted above, however, I didn't find anything (or, at least, anything I could understand!) about what to do with the first PI line when one doesn't want to fool with the programming. I'd assume (yeah, we all know) that it can simply be left unconnected and insulated, as if it required to be pulled down or something it would be both damned inconvenient to the wirer-upper, and also prone to causing customer complaints when the pulldown wasn't connected.
Although this PI/PO thing is an interesting diversion, not to mention some education, I'd venture that it's not involved in the problem.
Other things that occur - the DMX line should be terminated; if adding termination causes problems, then the problem is not the termination, it's something else that is broken, and the presence or absence of the terminator is just exposing that problem. Also, given that the PI/PO lines are directional (ie PI is in, and PO is out) I wonder if there is some directionality in the DMX flow too, implying DMX buffering per box, which would be, in a cheap Chinese box, highly unlikely, I grant you, but inbuilt buffering is not unheard of.
I'm still in the camp of corrupt signalling; is the CASbox adjacent to the chock blocks, or far, far away? I note the CASbox requires a class II power supply, so an ungrounded power supply.
TimSabre earlier suggested getting a multimeter out, and checking resistances and voltages. I support that suggestion.