Only hypothetical clocks can be supposed to be started or stopped instantaneously. Nothing happens without some time elapsing. An implication often suggested... especially in much mathematical modelling... is that when two events are deemed to be simultaneous they both happened together instantaneously. It is a mischievous fiction. Even the virtual timepieces in the imaginations of speculative cosmologists cannot be simultaneously started instantaneously. They can suppose that it can be done, but they cannot actually do it.

Since simultaneous simply means 'happening at the same time', it is not necessary that they both have beginning and end events that also happen at the same time. The seasonal event of spring, and the horticultural event of apple blossom flowering, can occur simultaneously as long as the calender period of one event is at least partially within the calender period of the other. It is not necessary that the both events stop and start at the same time. They are simultaneous during calender period when they are overlapping. Sexual participants in media presentations presume they are simulating simultaneous orgasms, as long as both performers fake ecstasy within a few plausibly noisy vocalizations.

In reality, the only pragmatic course of action is to judge two events to be simultaneous when the beginning-event of one clearly occurs before the end-event of the other. Determining whether or not two events started or finished simultaneously becomes a matter of deciding to what degree the two beginning-events, or the two end-events of each overlap. This determination is directly influenced by the choice of beginning and end events.
Suppose the arrival-event of the steam train was chosen to be that period of time between when the front of the engine entered the start of the platform to when the carriages became stationary. Suppose as well, that the time-tabled arrival-event of the 9:20 clock display was the period of time between when the display flipped from 9:19 to 9:20 and when the display flipped from 9:20 to 9:21. As long as the clock showed 9:20 at some stage between when the engine arrived at the station and the train had stopped, we can quite pragmatically claim that the events were simultaneous and that the train was 'right on time'.
If it is necessary to determine whether two events started simultaneously, then the beginning-events of each need to be checked to discover if they overlapped or not. Suppose an interval of time ΔT₁ is chosen to be the beginning of the extinction-of-the-dinosaurs event, and another interval ΔT₂ is chosen to be the beginning of an earth-comet-collision event. If ΔT₁ and ΔT₂ overlap at all, then assuming them to be simultaneous is a reasonable conclusion, and the one being the cause of the other a possibility. Which event might have 'caused' which would require some further reflection. On a grand geological scale, if layers of the shells of extinct sea creatures are located at the top of mountain ranges, we would be initially inclined to suppose that the extinction event and the tectonic event were not simultaneous, since the end of the first event most probably occurred well before the beginning of the second.

At the microcosmic end of the scale, relativistic considerations become paramount, and eventually, determining whether two beginning-events are simultaneous becomes essentially impossible in the context of the chaos of atomic uncertainties. Attempting to allow for time differences for information to travel from each of two separate events, or even from one event which initiates others, will eventually enforce the realization that the concept of an instantaneous event is an existential impossibility. There are no instantaneous events to be observed in the cosmos and there was no instantaneous event which initiated the cosmos .