After thinking about it some more, I recommend that each device on the power bus have a programmable I/V characteristic. This is especially important to avoid wasteful "sloshing" of large currents between energy storage devices.
Ideally, the storage devices would have a dead band in their I/V curves around a nominal operating point. They'd begin discharging only when the voltage falls below the lower edge of the dead band, and they'd begin charging at voltages above the dead band. Also, the I/V slopes outside of the dead band (i.e., their effective resistances) would be set to balance charging and discharging currents. The controllers would act like ballasting resistors, only they'd do it more efficiently.
You can do some useful things by reprogramming these I/V curves. For example, if you have both supercaps and conventional batteries, you might want to discharge the supercaps in an eclipse before the conventional batteries. You'd do this by setting a higher discharge kick-in voltage for the supercap modules than for the chemical batteries.
The IHU could also reprogram the I/V curves to run periodic tests on the individual storage modules. While normally you'd want to keep the batteries at equal charge states, you might want to intentionally charge and deplete one battery to measure its capacity while keeping the others full.
Some care will have to go into this to ensure that the module controllers always come up from a power failure with reasonable default I/V curves, but I think this is doable.
Phil