Six companies built one game. We kept track of what actually worked.

The challenge

Six organizations worked on BUSTA at once: the product owners, a core game studio, an exchange team, a smart contract shop, an interface team, and us. Work was split across Trello and Asana. Each team could do good work and the product could still fail, because no single party saw the whole thing.

Then it did fail in public. Starting 3 September 2021, players could not deposit or withdraw on the crash game for four days, and the manual workaround did not help. Every game instance shared one database, blockchain writes were running ten times a second, and the exchange launched the same week, adding traffic to servers nobody had measured.

What we did

1. One shared picture of the build

   From July 2021 we merged feature lists from every vendor into one view with an owner and a status per task. We wrote weekly status reports and recorded sprint demos on staging, so the product owner heard one account of progress instead of five.

2. Launch by checklist

   The mainnet cutover ran off a written list: the exploit fix merged, secrets moved to a managed store, the database rotated, the endpoint moved to its live address. Outside auditors tested the beta before launch, and their findings arrived on beta day rather than after go-live.

3. Write the failure down

   Three days into the outage we wrote an internal failed transactions report. It named the shared database, the write rate, a risky library upgrade and unanswered load-test requests. Its conclusion was that load testing would have prevented it.

4. Measure, then keep measuring

   We ran a formal load test in September and again in October with the same method. We also kept a production smoke-test workbook of nine steps per run, from wallet connect to withdrawal, against the live game rather than staging.

The result

The first load test set the safe ceiling at 120 concurrent users. Four weeks later the second put it at 350. The smoke-test workbook holds 120 production runs and 1,422 result cells: 1,312 passes, 9 failures and 101 not tested. The first run failed at step one because the wallet would not connect. Most runs after it are green.

A 44-scenario mainnet pass covered partner-token listing before any partner used it. When we checked in September 2026, busta.gg and the crash game were both still live.

What this means for your business

When several vendors build one system, someone has to own the true state of the whole thing, not just their slice. Know your capacity before you add users, and keep test records that show failures as plainly as passes. That record is what lets you fix problems instead of arguing about them.

This project was delivered by the Innovation Theory team, the Denver firm behind InTheory, before InTheory existed as its AI practice.