Since June of this year, we've noticed a spike in latency for DB queries routed through Supabase's Supavisor.

Direct DB queries will take a tens of milliseconds while queries through Supavisor will take a minimum of 200ms and up to 5s.

This is phenomena independent of the underlying query even `SELECT 1` through Supavisor can take up to 5s.

Has anyone else experience this before? Is this a known failure mode of Supabase's Supavisor?