Usually, as it has been said, this tag is used to test worlds which includes magic, monsters or something unusual in nature -that means also unusual in our world. So, a strict "reality check" is not possible, because:
- since reality doesn’t include magic, monsters or whatever, no one can predict what reality would be with it,
- even if it did, what we know about reality is never reality itself.
So, "reality-check" means more probably that you check that (the world) + (what you add in it, ie magic, monsters and so on) is coherent with the knowledge we have of our world.
Edit: a little bit more of explanation about it. This doesn’t mean that what you add (magic,…) is coherent with the knowledge about our world, but that a sum of all of that is coherent. Here is an example: if you imagine a world where magic can input free energy in matter, I’m ok to say it is not something known in our world, but the whole (magic of free energy + our world) is coherent because our knowledge can integrate this magic as if it were just something new. Actually, it is exactly the same with a new scientific theory: you can first check if this new theory is compatible with what is already known. We didn’t know that Earth is round before we could check it. And it is not incompatible with the knowledge "locally, the world is almost flat" (since locally, a ball is almost flat). So, this is what can be checked in reality check, I guess. Furthermore, as I said, you can’t check if an element is coherent with "a world", because you don’t know exactly what "a world" is. What you know is not "a world" or even "our world", but it is a set of theories about this world. So you can add theories about a new element in your set of knowledge about our world, and check if the sum is coherent. End of edit
But this implies two things:
- the thing you add in the world (magic, monsters…) should be described with enough properties to test if these properties are coherent considering what we know about the world,
- the state of knowledge itself must be specified, in the question or in the answer.
This last point is really important: since we don’t know everything about our world, it is important to write down what we know about it. For example, answers about FTL travel will be fondamentaly different in a relativity, Newtonian, or even more ancient context. Even historical questions could have different answers if you have the knowledge of the winner or of the loser of the war (maybe less true nowadays, but not that much).
It can be really interesting if your world is an ancient one: for example, if you are in the 14th century, you car more easily include FTL travel than in ours, since relativity was not part of the common knowledge at that time (even if it’s part of the reader’s one, of course).
So I’m pretty convinced in the fact that it is really important to explain the context and the hypotheses taken into account in the answer of a "reality-check" question.
To conclude, I would define "reality-check" as "Check if a world containing some more hypothesis is still compatible with a context of knowledge about our world, which must be specified in the question or in the answer."