Triage is triage: what the emergency shelf taught me about priority queues

Behind every pharmacy counter is an emergency shelf: the items that cannot stock out, cannot expire unnoticed, cannot wait for a reorder cycle. It's not the biggest shelf — it's the one where failure has consequences today. Software priority queues are the same structure, and most of them are managed worse than that shelf.

How the counter actually triages

By consequence, not by volume. The loudest customer doesn't move the emergency shelf; the drug whose absence harms does. Translation for software: incident severity is about blast radius, not about how many messages arrived or how senior the reporter is.

By time-criticality, honestly assessed. Some things can wait a day with no harm; some can't wait an hour. The counter distinguishes these instinctively. Backlogs often don't — everything labelled P2 forever, which means nothing is.

Review, not just reaction. The emergency shelf gets audited: what's on it, what expired, what stopped being critical. Most product backlogs never get audited — items age without consequence, and the queue becomes an archive.

What I run now

• a literal short list of things that "cannot stock out": payment path, auth, data integrity. Those get checked before any new work, every session
• consequence-based labels instead of effort-based ones: what breaks, for whom, how badly
• a scheduled audit where items must justify their position or leave the queue

The counter's wisdom is blunt: you can't do everything, so the only real question is what fails worst when you don't. Order by that, and the queue stops being a guilt list and becomes an instrument.