The hardened allocator doesn't wrap malloc but rather is the default malloc implementation on GrapheneOS with a per-app opt-out toggle. It makes no substantial difference for overall OS performance but there are definitely apps where it matters. OsmAnd works fine for most users without disabling it though.