I've just found some advices in arithmetic which shouldn't be there.
(dec (exp x)) and (expm1 x) are not equivalent. expm1 is more precise near 0.0 but it shouldn't be treated as drop in replacement in every case. The same for hypot (which catches overflows and underflow).
Also (long (+ x 0.5)) fails on overflow while (Math/round) not. Catching overflow can be beneficial in some cases so I would be careful.