There are many ways to test list_all_xxx, and this needs to be clear for vendors whose systems are under testing. I think the "induction" could be the best test strategy:
1. Considering starting the tests with no xxx (archetypes, opts, queries), then list_all_xxx should return zero items.
2. If there are N xxx (archetypes, opts, queries), and a new one is uploaded successfully, then list_all_xxx should return N+1 items, and the result should contain the item just added (match by xxx_id)
Note: for testing purposes I would suggest not to use the row_offset and rows_to_fetch parameters on the list_all_xxx services, since that difficult tests, for instance if there are X opts, add a new one and then want to check if the opt is there with list_all_opts, if by using the parameters we get less than or X opts, the new opt will not be on the result, instead need to change the offset to get the next page, then the opt will be there. I would like to avoid pagination on the conformance tests.