Float is "set of finite single-precision 32 bit IEEE 754 floating point numbers", unrestricted float is "set of all possible single-precision 32 bit IEEE 754 floating point numbers, finite and non-finite".

It seems obvious that NaN is not a float: it's not in the set of finite numbers. It also would appear that NaN is not a unrestricted float as it isn't a number. However, elsewhere in the spec we see the constant NaN exists as a unrestricted float.

We should clarify this (and also for the similar double case). There's also the question of whether the set contains 2^24-2 representations of NaN or a single representation of NaN (like the Number type in ES).