**Summary:** Google Chat users are experiencing an issue where they receive notifications for new messages, but the actual messages fail to appear in their chat...

What caused it: Google Chat's root cause analysis

Incident Report Summary On Friday, August 28, 2026, between 10:50 US/Pacific and 13:34 US/Pacific for a duration of 2 hours and 44 minutes, Google Chat users experienced delays in receiving real-time incoming messages within active conversation streams across web, desktop, and mobile applications. While users continued to receive notifications for new messages, the message contents did not automatically render in active conversation streams until the user manually refreshed the application or sent a reply message. To our valued Google Workspace customers, we are deeply sorry for the service interruption you experienced. We are taking immediate and decisive action to address the issue and strengthen our systems against future incidents. Root Cause The incident was triggered during a global database schema rollout. An incompatible database schema change caused database writes required for message delivery to start failing. Specifically, a mechanism for schema synchronization-designed to ensure high consistency between primary storage and live message delivery-encountered a rare race condition bug. This caused backend message dispatchers to attempt writes against an unready schema state, leading to write failures. Because live message delivery had an active dependency on these writes completing successfully before pushing events to user screens, the failure caused incoming messages to experience synchronization delays. Separate push notification services and core message storage operated independently and were unaffected, which is why users still received alerts for new messages while active conversation streams experienced delays. Remediation and Prevention The issue was detected through automated backend alerting on dispatch error rates, alongside corresponding customer reports. Google Engineering initiated parallel workstreams to restore service. * At 12:44 US/Pacific, Google Engineering directly mitigated the source of the issue by removing the dependency that live message delivery had on the failing database write. This allowed new incoming messages to bypass the failed write and immediately resume normal real-time delivery to user screens. * At 13:00 US/Pacific, Engineering halted and reverted the database schema change to its previous verified stable version, restoring full database compatibility. * By 13:34 US/Pacific, as queued messages were processed, real-time message synchronization delays returned to 0% across all customer domains. Engineering ve...

Read the full postmortem from Google Chat