In very general terms, even with no assumption about priorities, the decision about which one to service is irrelevant. Whichever one goes first the other one must wait until the first one is complete, then take its turn. In some processors it is possible to allow nested interrupts. This should be avoided until you are really comfortable with the hardware and the system you are trying this on, and have a good reason for wanting to do it.
Under the assumption that interrupt processing time plus interrupt latency time is very much less than the interrupt arrival rate, then processing never falls behind. This is probably the most important reason why interrupt routines should be short and fast. It is almost always a mistake to perform lengthy and non-deterministic amounts of processing in an interrupt routine.