The code in systemDictionary computes both hashcodes when they're the same, so add some asserts and only compute one of them.

The system dictionary computes the hashcode on the name and index outside the lookup code which might have been an optimization, since there are a few lookups. It makes the code messy. Maybe the hashcode and index computation should be inside the placeholder()->find_and_add(), find_and_remove(), get_entry() functions etc instead.