Salesforce experienced a major service disruption on September 16, affecting customers across multiple regions and causing login issues, intermittent errors, delayed services and difficulties submitting support cases.
The outage began at around 3:50 a.m. EDT and was officially resolved shortly before 3 p.m., following several hours of monitoring. Salesforce initially linked the disruption to an external dependency failure involving its legacy login infrastructure. Increased load on a core system component subsequently reduced its capacity to process requests.
The incident occurred while Salesforce was hosting its flagship Dreamforce event, but its significance extends beyond the timing. The disruption highlights how cloud platforms can contain deeply interconnected dependencies that may create widespread consequences when one critical component fails.
Recovery took several hours
Salesforce found that requests were becoming stalled while waiting for an internal login service that was consuming available server resources.
Engineers responded by restricting API endpoints, carrying out rolling restarts and deploying fixes across regions. Some customers began regaining access during the morning, but recovery was uneven. Certain instances required additional intervention, while some scheduled jobs continued to experience problems even after services were restored.
A subset of Hyperforce instances was among the last to recover. Salesforce had implemented mitigations across its instances by approximately 11:39 a.m. and continued monitoring before declaring the incident resolved.
The company said it would conduct a detailed investigation into the technical trigger, underlying cause and preventive measures.
Restoring access is only the first step
For organizations using Salesforce as a system of record, a prolonged outage can create problems beyond temporary loss of access.
Transactions may be delayed, integrations can accumulate retries and queues, scheduled processes may fail and records can temporarily become inconsistent. Customer interactions handled through alternative channels during an outage may also fail to reach connected systems at the expected time.
This makes post-outage reconciliation essential. Enterprises should verify APIs, integrations, scheduled jobs, workflows, authentication processes and downstream systems rather than assuming that restored login access means every business process has returned to normal.
Security teams should also review authentication activity, privileged access, integration credentials and emergency changes introduced during recovery.
Cloud modernization does not remove dependency risk
The incident also demonstrates why legacy technology cannot be assessed simply by its age.
An older authentication service can remain a critical component within an otherwise modern cloud architecture. Its importance depends on where it sits within the dependency network, how many services rely on it and how effectively failures are isolated.
For enterprise architects, dependency concentration, failure isolation, recovery paths, graceful degradation and blast radius should therefore form part of modernization strategies.
The critical question is whether a failure can remain contained within one component or cascade into broader platform disruption.
AI and workforce factors remain unconfirmed
There has been speculation about whether AI-assisted software development could contribute to large-scale technology failures. However, there is currently no confirmed evidence establishing that AI caused the Salesforce outage.
Workforce reductions at Salesforce have also been raised as a possible factor affecting operational resilience, although no direct link to this incident has been established.
For enterprises, the central takeaway is architectural: cloud adoption can modernize infrastructure while leaving critical dependencies intact. Mapping those dependencies, strengthening isolation and testing recovery mechanisms under failure conditions remain essential to building resilient digital systems.
