>"How come bias T's just use the case?"
Because the skin effect is useful in RF shielding, because it just happens to RF, not DC. You want DC on your power return (the ground) and ALL the Rf to remain inside the box and on the RF lines.
BTW, I am having a hell of a time right now with RFI on an RPI hat - which I am thinking of abandoning in favor of short ugly construction.
If you want a great way to learn how RF shielding works, I suggest building some LNAs using MMIC chips. Try out different construction methods, cases, RF interconnects, DC interconnects, DC blocks, bias-tees, and via styles. The fact that LNAs amplify amplifies your learning!
Using SMT makes it a lot easier probably because it eliminates a lot of potential complex effects from the picture. Less is more in that respect.
They have some pretty neat feed-throughs now that claim to incorporate a whole RF filter inside. Not just a capacitor.
The standard-shaped feed-through ones work well.
That should be enough in most situations. If ts not maybe add some more bypassing and L outside the box.
But here I am still struggling with an RFI situation that should be addressable. Need to throw a bunch more caps and ferrite beads on there. Bypass the power lines with those 3 terminal 2d feed through caps for SMD. Or put the RPI and GPS in a partitioned screened metal box. (because of the heat) Or dump the hat.