What copying these benchmarks into that account would do. Writes nothing.

One endpoint for many sets rather than one call each, because the expensive part is resolving the target: every single-set dry run opened its own connection to the GetPlace2 production database and looked up the SAME account again. Selecting fifty rows meant fifty connections to somebody else's production. Here the account is resolved once and the answer is computed per set — the access report genuinely differs, because the brands differ, but the account does not.