This post voluntarily spells things out in way more detail than is necessary. Don't assume I'm making fun of you (or anybody), I just don't know what level of understanding of my post you're at, so I'm providing full detail.
I (and probably many others) keep track of the closest defibrillator found by remembering its distance to me in a variable. I update that variable as I evaluate other defibrillators. To keep the code nice and clean, that variable has a default value before it gets pitted against the defibrillators. (The common alternative is to single out one of them, say the first, but there are various reasons why it could be considered not as elegant)
The only safe practice is to use a quantity garanteed to be higher than all others to be considered in the problem. Like positive infinity. Unfortunately in many programming languages, infinity is not convenient to use as a value. So it's quite natural to simply use a very high value.
I happened to use the earth radius. (I haven't investigated it so much more deeply, but the symptoms are that) it worked in the IDE tests, yet failed in validation test #1. Using 1000 or 10000 (don't remember) times the earth radius happened to make the test pass.
So my suspicion is that the test case uses coordinates way out of range for Montpellier. Surprising in context, but why not. Another explanation would be that the distance formula provided in the problem could be faulty. I haven't checked anything more thoroughly, as I did validate the problem eventually. I was simply dropping a post here to give others who made the same "mistake" as me a chance to spare some debugging time.