A ledger names each of the server's mutations with scenarios that run it
against shared, realistic data and check, for every strategy and both
users, that a changed full read always changes the listContentVersions
answer; mutations that cannot touch a full read give the reason instead.
A new mutation without an entry fails the test.