So that the micro VM implementation does not have to record the IDs and
names for many local things, such as instruction results and clauses.
These things are numerous, but the API cannot do anything on them. Not
forcing the micro VM to record their IDs/names can save much space.