I don't agree with the assumptions. My experience tells that the sum of module power dissipation (IGBT + diode, forward + switching losses) won't change more than e.g. 10 percent when switching between motoric and generatoric mode. The detail behaviour is bit more complicated because also modulation index (actual motor voltage) and phase angle affects the individual losses.
We can also assume that grid and motor inverter are (at least partly) sharing a common heatsink, in so far the module overtemperature will be further balanced.
In case of doubt, it's of course better to ask ABB about possible limitations in regenerative mode.