The combination of stronger type systems (especially dependent typing), less error-prone design (no manual memory management, no mutation, no global state, no loops/off-by-one errors), separated side effects (crashing during computation won't break things, less places for outside interference/external failure, etc.), and better error handling (no null, usually no exceptions meaning you have to encode failure into the return value itself without losing type information) solves many common bugs in imperative code.
Any examples of how it can help out?