This underscores the notion that functional programming is ultimately about immutability, and all the idioms expected by a truly functional language are really a result of how it handles state more than anything. A closure is usually expected to close over values, and the idea of a "variable" doesn't really mean the same in a truly functional language, which is where all the confusion and the pain points stem from.
Yes, if you come from functional languages then you might expect to close over values, but I don't think that in general "a closure is expected to close over values". In particular, both Scheme and Common Lisp close over variables, not values.
JavaScript also closes over variables, so it's not just "obscure" languages that behave this way.
Java only allows you to close over variables that are effectively final, to remove confusion on both sides (e.g. why doesn't the value in my closure change when I manipulate the original variable? vs. why does my the value in my closure change when I manipulate the original variable?)
