I would assume the CMU has a connection to each 3P cell group (98 connections total)Balancing will be done in the same unit that monitors cell voltages, as it has connections to each cell.
Bear in mind that balancing will be making small changes to adjust cell voltage, and can do this fairly slowly, so the power dissipation isn't likely to be very high. There are some pics of some BMS modules on the openinverter forum - looks like a bunch of transistors and resistors, e.g.
Kia PHEV : https://openinverter.org/forum/viewtopic.php?f=13&t=1179
Zoe : https://openinverter.org/forum/viewtopic.php?f=13&t=1179
Curious just how much of an imbalance between the individual cells in that group would trigger an error code. If one cell were completely dead, the other 2 could pick up the slack even with a little help from adjacent healthy groups. A condition over time that would be more distinguishable, but in the short term may be overlooked.
