With #87084 we identified that multi-tenant configurations are significantly slower that single-tenant configurations when running TPC-H. The theory is that this is due to the inability to push certain SQL operations down into the KV layer. This however, would only be an issue for separate-process MT configurations. For the shared-process UA implementation, we should validate that there is no discernable performance difference when running TPC-H.

Epic CRDB-23559

Jira issue: CRDB-19444