Why does radiometric dating tell us the ages of rocks? I understand well how radiometric dating works in theory. 
In practice, I'm not sure how one would calibrate this for rocks. I've always been told that we can determine rocks' formation time via radiometric dating. This implies that something about the radioisotope abundances are known (or assumed) at the time of formation. 
Teachers have said that rocks' "internal clocks" reset when they form. Suppose you take a 1 Gya rock, melt it in a furnace, and then let it re-freeze. According to what I've heard, the rock's "internal clock" has now reset, which implies that its radiometric abundances are the same as when it first formed. This doesn't make sense: if, over its 1 Gyr lifetime, it lost 99% of isotope A, it won't suddenly get the original abundances of that isotope back through the process of melting and freezing. How, then, would one be able to distinguish its date of formation? 
 A: I am by no means an expert in the many radiometric means of dating rock, but one of the systems often used (and the one I believe your teacher was speaking of) is potassium-argon dating. Rocks with potassium content will contain some primordial level of radioactive K-40 at the time of their formation. That stuff has a half-life of 1.2 billion years so it is well suited to dating stuff over geological time spans.
Potassium-40 is beta active and decays to argon-40. Which is a gas. A noble gas. So it doesn't bind chemically to other atoms in it's environment and will escape if not confined.
If that decay happens in a solid context the argon is trapped, and a scientist can put a small sample of the potassium bearing mineral from the rock through a mass spectrometer to measure the ratio of K-40 to Ar-40 present, and from that determine how long it has been since the rock was last in liquid form.
Because gaseous argon moves quite easily through molten rock and will escape from the immediate neighborhood of its potassium progenitors.

Other radiometric dating schemes for geological samples may not have the property where the daughter isotope is a noble gas.
