Knowing the DSP latency and connectivity were the keys for me pushing the Modular Mixer as far as I could go with it (there is a chart that is an expanded version of this:
http://www.hitfoundry.com/issue_07/xitemast.htm , that I can't seem to find atm, but lists latency between each DSP as well..basically slot 10-18 is fastest to I/O, intra-slot is a couple of samples, while inter-slot can add a few extra cycles) . I learned some decent lessons from loading a standard STM24...it is spread out across the 1st DSP of each slot. So that's about 8 channels on each DSP 7-10. Using 16 channel Stems and such create difficult routing issues between DSPs. Moving to smaller mixes, and carefully placing them helped me get up to a larger format mixer, with plenty of aux, busses, and even gating/comp on all channels. Reducing the size of the mixers also leaves some space on the DSPs for eq and effects on each channel.