i am just another OV user, but not sure if this doc answers either one of your questions regarding semantic labeling (but feel free to take a look in case there are some relevance):
https://docs.omniverse.nvidia.com/extensions/latest/ext_replicator/semantics_schema_editor.html
my understanding is that the warehouse level uses SimReady assets as reference. and each geom prim already has predefined semantics. furthermore, SimReady library assets that were shipped with/used in OV apps were defined by the general guide found here - Semantic Labeling — Omniverse SimReady latest documentation (the bottom URL should direct you to the spec doc)
i am not too sure how replicator allocates the ID numbers, though.