I think while all these are probably all good guesses as to changes on the new Mac Pro, none of them really produce the "mini" Mac Pro a lot of people are hoping for.
I wasn't trying to motivate a "mini" or even "mid-tower" Mac Pro. These all modern functional demands that the current case is oblivious to because it is motivated by legacy design constraints/parameters.
GPUs and high end I/O cards are just as hot if not hotter than CPUs these days. that has nothing to due to with going "mini". Non-ODD storage devices don't need to project a tray through the size of the box. Again external impact but nothing really having to do with size. Even if swap for external drive sleds ( reuse and modify old XServe sled ideas or adopting some SATA USM standards for removable drives ) it would have an impact on the external design. Same for multiple antennas... big or small box still need to place them so that useful ( future airport extreme and large tower Mac Pro would both need multiple antennas for things like 802.11ac to be effective. )
I'm mixed on the handles. They're pretty handy (hah) if you've ever had the experience of lugging around a Dell workstation.
The handles aren't really primarily handles. It is a "footer" to raise the box up off the ground. The same set on top is just as much driven by Apple's OCD symmetric design principles as high utility as something oriented to the human hand. They are sharp and flat because that makes for a great foot, not because it is substantially easier to carry.
The 7600/5600/ Dell boxes have handles only on the top of the sides. HP's boxes are similar. The design difference is that they are design not to stretch up the height of the box. Apple could still keep the symmetric footer/handle design. It just doesn't need to be quite so high (keep all the external & internal symmetry and just expand the dimension of depth around its central axis.). Or it doesn't need to be permanently attached ( make the core simple rectangular box transformable. ). There are other designs out there that meet the design parameters better than the current one does.
In fact it is somewhat indicative that the Mac Pro was taller in part to help segment the market between it and the XServe to a further extent. If promoting that segmentation went into the core design, it is completely unmotivated now that the XServe is gone. That doesn't mean the Mac Pro needs to fit all of the XServe functionality, just that they don't need to overtly minimize overlaps.
The Mac Pro would get substantially smaller if used substantially more limited parts. For example dropping back to a Xeon E3 with a much narrower TDP. That in turn also chops down PCI-e lane budget to roughly a single slot. Cap that slot to a sub 230W power budget and then can drop the power supply and PCI-e card zone cooling requirements. Those would do a major reduction. Substitute 2.5" drives for 3.5" and you could easily do a 1/3 shrink of all of the Mac Pro's thermal zones without much drama. Tossing the ODDs provides room for front I/O sockets and opens room for an antenna or two.
The issue though is that you cut the performance significantly. Capped at 4 x86 cores. Lost 20-60 PCI-e v3.0 lanes of I/O bandwidth. (an amount 5-15 times as large as Thunderbolt's. ). I can see utility of something that is more oriented to being a SAN client node. Lots of the space consuming I/O would be outside the box and shared. ( shared mobile/desktop with TB and shared storage through something like 10GbE. ) That smaller and priced close to the edge of the $2000 border would probably do much better than the single package current designs.