The primary cost benefit of orbital data centers is the availability of efficient, around-the-clock solar power. Any time spent idling for heat reasons directly eats into the profitability of the concept so there's no reason* to design a system that isn't optimized to operate in a steady state. You would instead just put fewer GPUs on it or increase the size of your panels and radiators. The only question is whether or not that steady state is cost effective.
*There are reasons, but none that justify very large scale compute clusters in space
*There are reasons, but none that justify very large scale compute clusters in space