Creating a full
GIN (payload) index on a heavy JSONB column bloats write amplification when you only ever filter by payload->>'event_type'. A targeted B-tree expression index on ((payload->>'event_type')) is a fraction of the disk size and much faster to update on…