Who's using our parallels right now?

Two enterprise customers with hard capacity ceilings asked the same question in different accents: which teams are consuming our parallel test slots right now, and is the queue fair? This note maps the design space for real-time usage visibility — an API, a dashboard, or both.

Problem

At fixed parallel capacity, one team's heavy build is another team's queue. The customers asking — a top-tier bank managing a hard cap, a fintech platform against a licensing ceiling — needed attribution (who triggered what), live queue pressure, and a lever to act, none of which existed at the granularity operations teams work at.

Where it leads

This is research work — it ends in open threads, not outcomes. They're below, stated plainly.

The use-case matrix drove everything: real-time usage by user and team, queue pressure when limits are hit, attribution of any running build, the ability to stop or de-prioritize, and history for fairness analysis. Against those, the delivery options split cleanly:

No single shape covers the matrix — which is the finding. Enterprises with ops tooling want the API; the lever to act wants the dashboard.

The recommendation under exploration is a sequence, not a choice: ship the API as the source of truth, then surface the same data on-demand inside the dashboard — fetched at the moment of request, honestly timestamped, no pretense of a streaming feed. The concept work exists as early explorations; the note stays open until an enterprise ops team has torn it apart.

Open threads

The phased bet — API first, dashboard reading from the same API second — is drafted but unvalidated: does an ops team accept 'on request' freshness, or is anything short of live a broken promise?
Fairness is the unstated feature: usage attribution turns capacity from an infrastructure fact into an organizational negotiation. The design will be quoted in someone's budget meeting.
De-prioritizing a running build is a power tool with blast radius — who may wield it, and what the audit trail shows, is a governance design problem stapled to a UI one.
Minute-level granularity has real infrastructure cost; the design honestly needs less than 'real-time' in most flows, and saying so out loud may be the cheapest feature.

The full walkthrough — screens, numbers, names — happens in conversation. Start one →