For our surface stim paradigm, the longest pulse we use is 500 usec (with a 2500 usec recovery phase). So if we assume a 3000 usec (3 ms) total pulse time, at 100 Hz we could conceivably do 3 channels with this setup, correct? Or, if we go down to 2 ms total pulse duration, including recovery, we could easily do 4 channels at 100 Hz. At lower frequency, we could then go to more channels, i.e., at 50 Hz, we could do 8 channels with 2 ms total pulse duration. Is my thinking correct here?

If we allow 3ms for each stimulation pulse and a 1ms gap between them to allow the D188 to switch, then this give a total time of 12ms. This should mean each site can be stimulated ~83Hz. With the pulse duration reduced to 2ms and retaining the 1ms interval, this increases to 111Hz.
It is important that as well as the total time spent stimulating, they also allow for the D188 relays to fully switch channels. The 1ms may be a little conservative, but they could experiment with timings to get an idea of how fast they can do this with their particular control method.