For this iterator the BLI_Iterator was used, not BLI_Iterator.current unlike rna_Depsgraph_ids_get.

This is working as intended, and doesn't depend on the location of BLI_Iterator in RNA_DepsgraphIterator, double checked that putting py_instance first works.

Nope, not done :P

Oooooh right, all members of the DepsgraphObjectInstance are only accessors, so the only thing that matters is that they all agree on the type of ptr.data!
Sorry about that.

This could be de-duplicated with the same code above maybe? E.g. in a pyrna_prop_collection_subscript_end ?

Thought the re-wording addressed your comment, re-worked again for better clarity.

Agree with the goal of de-duplicating boiler plate.
An issue with this is it moves Py_DECREF into a function, which can make reference counting a bit more difficult to follow.

if (found) {
  if (result && (pyrna_prop_collection_subscript_is_valid_or_error(result) == -1)) {
    Py_DECREF(result);
    result = NULL; /* The exception has been set. */
  }
  return result;
}

Could be replaced with:

if (found) {
  pyrna_prop_collection_subscript_is_valid_or_error_and_clear(&result);
  return result;
}