War games: how we built Kraken to handle 10x the load
Kraken introduced automated load testing to handle extreme trading spikes after failures during Bitcoin’s July 2025 rally and a token sale in October 2025.
The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.
In July 2025, Bitcoin’s surge past $120,000 triggered errors for 135,500 clients on Kraken, with 11% of requests failing at peak times. Three months later, Kraken Launch’s token offering caused a 60-80% drop in active sessions for two and a half hours, with persistent request failures during spikes. These incidents exposed vulnerabilities in handling sudden, extreme load. The company acknowledged that platform stability during such events defines its reliability, not average performance.
To address these failures, Kraken implemented full-scale load tests called "war games," simulating trading activity to identify system constraints before real market stress occurs. The tests operate on the company’s schedule, allowing engineers to stress-test production systems and fix weaknesses proactively. This approach replaced reactive measures with a standardized requirement: the platform must handle at least double the heaviest recorded live trading load on demand.
The new testing framework proved critical in February 2026, when a sharp Bitcoin selloff drove trading volumes to all-time highs. Despite the extreme conditions, Kraken reported no service incidents, estimating that a full outage would have cost $719,000 per hour. The platform’s peak failure rate dropped from 11% to 0.111% between July 2025 and June 2026, while handling over ten times the load.
Kraken now permanently tests against twice the most recent record trading load, ensuring the bar rises with market activity. Each new client-driven record becomes the benchmark for future tests, reinforcing the platform’s resilience. The company plans to publish a technical companion post detailing the testing methodology, failures encountered, and lessons learned during the process.