Review Scope Discipline
Stating what an architecture board does not review, which is what determines whether it stays useful or becomes a queue.
Every architecture review function drifts towards reviewing more. Each addition is individually justifiable, and the aggregate is a body that meets fortnightly, has a six-week backlog, and is routed around by anyone with a deadline.
The corrective is to state the exclusions as explicitly as the inclusions. A board reviews decisions that are expensive to reverse, cross more than one team, deviate from the technology radar, or introduce a new capability to the estate. It does not review conformance to standards — that is automated — nor implementation choices within an agreed boundary, nor anything already covered by a published decision.
Two further disciplines keep it working. The board should decide, in the session, rather than requesting more information — a body whose usual output is "come back with more detail" has a longer effective backlog than its calendar suggests. And it should publish its decisions as records so that a question answered once does not need to be asked again by the next team.
The measure worth tracking is time from request to decision. When it exceeds a sprint, teams stop asking, and the board's actual coverage falls to whatever arrives voluntarily.