The Broadway Drop Shop
Limited-edition “drops” gated by a custom virtual-queue engine.
Summary
A fully custom, original — not white-labeled — online store designed specifically for limited-quantity drop sales, with a from-scratch virtual-queue engine that admits buyers fairly under heavy simultaneous load. I built it as Co-Founder and solo technical designer and developer. It’s the same launch-day concurrency problem behind game and console drops, ticket on-sales, and limited releases: everyone arrives in the same second and the queue has to stay fair.
The Problem
High-demand drops melt naive checkout flows: everyone arrives at once, sessions break mid-purchase, and fairness matters. It needed a real queue, not a best-effort one.
Approach
- Built a Redis-backed queue engine with position tracking, timed admission, and minute-level cron reaping of expired entries.
- Wrote a custom Lua admission script that maintains the ordered queue while limiting how many buyers may enter checkout simultaneously.
- Designed a dual session model so a queue position survives a mid-drop login.
- Covered the full order lifecycle with drop attribution.
Architecture
- 01Next.js 16 + React 19 + TypeScript; Panda CSS and Framer Motion.
- 02Supabase (PostgreSQL + Auth) for data and identity.
- 03Upstash Redis queue engine driven by a custom Lua admission script: position tracking, timed admission, cron reaping.
- 04Dual session: HMAC anonymous cookie linked to a Supabase JWT so position survives login.
- 05Resend + EmailOctopus for lifecycle email.
Key Tradeoffs
Custom Redis queue over an off-the-shelf waiting room
Drop fairness and timed admission needed precise control of position and reaping that hosted waiting rooms don’t expose.
A Lua script for admission, not application code
Admitting the next buyer has to be atomic against Redis; running the decision inside Redis removes the race an application-level read-then-write would leave open.
HMAC cookie + Supabase JWT dual session
Buyers join anonymously then log in mid-drop; binding the two keeps their hard-won position intact instead of dropping them to the back.
Stack
Outcome
A fair, resilient drop platform — custom queue, capped concurrent checkout, sessions that survive login — handling high-traffic simultaneous releases.