On July 21, 2026, we covered a cooling failure at a Google Cloud facility in the Netherlands that knocked europe-west4 services offline for around 15 hours. Six days later, Google shipped the feature that would have materially changed the shape of that incident for Cloud Run workloads.
What Service Health actually does
Cloud Run’s Service Health feature reached general availability in late July 2026, and it automates something teams have historically had to wire up themselves with custom health checks and DNS failover logic. Individual container instances run readiness probes, and Cloud Run aggregates those results to determine the overall health status of each regional deployment. When a region’s health status degrades, incoming traffic through a global external Application Load Balancer or a cross-region internal Application Load Balancer is automatically routed to serverless network endpoint groups in a healthy region instead. Setup is described as two clicks, and there is no additional charge beyond the CPU and memory the readiness probes themselves consume, which removes the usual cost objection to running proper multi-region resilience.
Why the timing is the actual story
Individually, an availability feature shipping is not news. What makes it worth flagging is the six-day gap between a real, hours-long europe-west4 outage and Google turning on automated protection against exactly that failure mode for one of its most widely used compute products. It is a useful, concrete reminder that cloud provider resilience tooling tends to catch up to specific outages rather than anticipate them, and that any workload running Cloud Run without multi-region deployment and without this feature configured was, until very recently, exposed to precisely the kind of regional facility failure that took europe-west4 down for most of a working day.
What this means for teams running on Cloud Run
If your Cloud Run services are deployed to a single Google Cloud region, this is a low-effort, no-additional-cost way to close that gap, provided your architecture can support a genuinely multi-region deployment with a global load balancer in front of it, which is not automatic and needs actual design work around session state, data replication, and where the actual data your service depends on lives. For European teams that specifically felt the europe-west4 outage, or that are simply reviewing single-region dependencies as 2026’s run of cloud provider incidents continues, this is a concrete, low-cost item to move up the backlog rather than leave as a someday improvement.
If your organisation runs on Cloud Run, AWS, Azure, or Hetzner and needs help designing genuine multi-region resilience, not just a checkbox failover feature turned on without the architecture to back it, contact Excello Digital. We help European teams build cloud infrastructure that survives a bad day at one data centre.