It's not a matter of multiple inertial frames--there is only one frame in this problem. Instead it's a matter of realizing that in order to interpret raw observations, one must account for light travel time.
Generally, in relativity thought experiments, one imagines an inertial frame as having zillions of observers (each with his own synchronized clock) located at every possible position. Thus when the lights flash, there will be a co-located observer able to confirm that the lights all flashed at the same time. (The various observers just compare their clock readings later.) But if an observer sees some event happen that is remote from him, he must subtract the light travel time from the time when he saw the light to find the actual time (according to clocks in his frame) that the event happened.
Of course, observers in a different frame (moving with respect to the bulbs) will disagree that the bulbs flashed simultaneously.