The sibling service follows the house style; this one never adopted it. ## Evidence - `views/application.py:86,271-279` — the dashboard list issues **1+2N** queries. - `models/application.py:165-173` — `needs_my_decision` costs roughly **four queries per application**. - Zero `select_related` / `prefetch_related` calls in the repo. This gets worse exactly as the portal gets used: the dashboard is the first screen a reviewer sees, and it scales with their workload. ## Scope Add the query optimisation across the list and dashboard paths, with `assertNumQueries` tests so it does not regress. **Sizing:** Weight 3 — multiple views and models, and the tests are the point. --- *Filed 2026-07-30 from a repo-wide audit pass over the marinerg estate.*