Following the 2024 incident where UniSuper's data was erroneously deleted by Google Cloud Platform (GCP), another severe configuration error has struck a major client. On May 19, 2026, GCP's automated systems mistakenly suspended the production account of Railway.com, a prominent Platform-as-a-Service (PaaS) provider, resulting in a total service outage lasting approximately eight hours.
According to Railway's official incident report, the suspension occurred at 22:10 UTC on May 19th. This action immediately severed Railway's access to critical GCP infrastructure, disabling its control panel, API services, and core networking components. Although Railway managed to contact their GCP account manager and restore the account by 22:29 UTC, the recovery of actual services was far more complex. Compute instances, persistent disks, and network configurations had to be restored piece by piece, with the incident finally being resolved at 07:58 UTC the following day.
In response to the outage, Railway announced a significant strategic shift to mitigate future risks. The company plans to reduce its reliance on Google Cloud by removing GCP from its "hot path" of production operations. Moving forward, GCP will be relegated to a backup and failover role. This incident underscores the ongoing reliability concerns regarding GCP's automated enforcement systems and highlights the risks of single-cloud dependency for mission-critical infrastructure.
