Bunsen confirmed that the IIcx PowerCache adapter is passive. I had one and can't find it, but that's what I recalled.
Results so far:
AFAIK, every Mac that has an implementation of Slot ID $E in any form except the LCIII and IIci require active adapters.
IIci doesn't count because its implementation of cache memory space (possibly NuBus SuperSlot memory space is involved?) and $E implemented in NuBus are compatible by (definition) design.
LCIII is a special case because its PDS is an oddball hardware implementation of a full '030 PDS connector at Slot ID $E. Installation of the PowerCache adapter in its PDS precludes any other use of $E, hence no possible conflict and its passive adapter.
Members of the Mac II series have NuBus implemented in Slots $9-$E. The IIci is an oddball, with slots $C-$E implemented in NuBus.
IIcx follows the pattern of its big brother the IIx, implementing the first of its three NuBus slots at $9-$B. No implementation of Slot ID $E is made as a physical slot or a pseudoslot in its design. No possible conflict there means only a passive adapter is required for the PowerCache.
From this, I think it's pretty safe to say that unraveling the mess of the SE/30 and IIsi video implementation at $E as compared to the IIci's implementation of NuBus at $E would be the way to begin understanding the workings of their slot adaptation mechanism. First guess is that it has to do with where cache is addressed in slot space memory?
Because of the way QuickDraw and the Slot Manager work, PDS video cards typically want to be seen by software as NuBus. And video cards want to be in the highest numbered slot, because the interrupt latency is lower at the top end of the range.
Due to the way NuBus slots map to memory addresses, slot $F isn't possible (the slot $F super space is where the other slots' non-super space goes), and so slot $E is your most efficient place to park a video card. That's why the SE/30's on-board video pretends to be slot $E. The IIci and LC series on-board video mostly doesn't pretend to be NuBus at all, and isn't subject to those restrictions (they report interrupts in an entirely different way, for one thing).
Yep, know how Slot Manager treats Slots and Pseudoslots.
Synopsis of information on pp. 410 & 412 of GttMFH2e*
IIci Video is addressed as PseudoSlot $0 (the same as the rest of the Mac) while video memory is accessed in NuBus memory space $B.
I've not seen a DevNote on the LC or info on this in the Bound PaperDocs. It's probably similar to the IIci as you said. It's curious that the LCIII PDS is located at $E.
I've heard much about the importance of slot positions for video cards, but never actually seen it documented in print or in PDF about hardware, DevNotes included. Is that information from Inside Macintosh documentation?
The additional info about the IIcx PowerCache adapter plugging directly into the CPU socket is much appreciated, the assumption (ass me) that this would be known to all readers was my bad. I might further clarify that no PDS exists in the IIcx, hence the adapter plugging directly into the 68030 socket.
This would be analogous to the "socketed" PowerCache for the SE/30. Now I wonder if there is any adaptation on the Logic Board of that version of the PowerCache Accelerator?
Another possible point of differentiation in the case of that particular implementation and much food for thought there.