I have just updated the details section to update some of the changes made recently:

1. Change from a "one-MCU per channel" design to a "single MCU controlling each channel" design - this will simplify firmware uploading (you can upload firmware over USB, and you only need to update it once rather than N times for N channels).
2. Change from an XMega to an ATMega32u4 - this will simplify programming (more people have access to ISP programming gear vs. PDI) and development (since I already know the ATMega family inside and out).
3. Adding the idea of programmatic control over USB (set voltage / current setpoints for each channel) as well as logging actual voltage / current readings. This will be done over Raw HID for driverless operation on all platforms.
4. Removing the adjustable A/C output requirement. Trying to drive an A/C waveform at reasonable resolution (at least 256 samples) and frequency (at least 60Hz) over I2C is going to be too slow and processor intensive. Due to the MCU change, using a built-in DAC driven by DMA is no longer an option. However, all is not lost - I am still planning on bringing out the raw A/C output from the transformer, which means that I will have 18VAC at a high amperage available if needed.
5. It makes extensive use of screw terminals for interconnecting the various boards
6. When routing, I am targeting 1oz copper PCBs, which means the traces need to be thicker than usual. All traces are sufficient for a max of 8A / channel with the entire system limited by the transformer (to a reasonable point, of course), EXCEPT for the traces connecting the OPA549 to the V+ / V- inputs. For this section, if you plan on using more than about 6.5A you will need to solder on some thick jumper wires to carry more current.

In addition, I have the first draft of the schematic completed, along with a routed board coming in at just under 5x10cm (sans the UBBB32u4 controller board, which will be panelized in later). See both below: