It works because a question opens a piece and an explanation closes it. Dropping an audience into a scene already in motion creates an immediate gap (who, why, what just happened), and the mild disorientation is the mechanism. People will sit with confusion for a long time. They will not sit with indifference at all.
The technique has a debt attached. You have to pay the context back, and payments that arrive too late feel withheld while payments that arrive too early waste the opening. The usual rhythm is to resolve who and where quickly, and to hold why until it can do structural work.
It is now close to mandatory in short-form video, where the first second decides everything. A piece that opens on a person introducing themselves has spent its best second on the least interesting information it contains.
What does in medias res mean?
Beginning in the middle of the action, with context supplied afterwards. The audience is oriented by consequences, which creates an immediate question.
Is starting in the middle always better?
Nearly always the stronger open, and introductions can wait. The technique creates a debt (the context has to be paid back), so it fails when the answer arrives too late to matter or too early to have earned anything.