Salesforce hit by global outage on day two of Dreamforce 2026

What happened

Salesforce suffered a global service disruption on September 16, 2026, the second day of its Dreamforce conference, leaving customers around the world unable to log in to the platform.

According to an account published on dev.to, the disruption began at approximately 07:50 UTC and affected Salesforce's Core Service. The author, who says the timeline is drawn from Salesforce status updates and press coverage at the time, reports that hundreds of instances were impacted across every region, including the United States, the United Kingdom, Germany, France, Japan and India.

Symptoms included severe delays, intermittent errors and failed logins, with some customers struggling even to submit new support cases.

Preliminary cause and recovery

The dev.to post says Salesforce's initial explanation pointed to requests stalling on an internal login service until available capacity was exhausted. The author notes this cause is preliminary and directs readers to Salesforce's Trust status page for the official root cause analysis once it is published.

A fix was reportedly validated and rolled out late in the morning UTC. Recovery was uneven, according to the post: some instances came back more slowly than others, and after service returned, some customers found that scheduled jobs had not run as expected.

Awkward timing

The outage landed in the middle of Dreamforce, Salesforce's largest annual event, held in San Francisco. According to the dev.to author, the on-stage programme centred on the company's AI ambitions — Agentforce, "AIforce" and "Headless 360" — while conference attendees and customers elsewhere faced an unavailable platform.

A familiar failure mode

The post places the incident within a longer pattern, listing major Salesforce outages in 2019, 2021, 2023 and 2025, attributed variously to faulty permission scripts and DNS changes.

The author's wider argument targets developers: the standard Apex workflow — writing code in an editor, deploying to a scratch org or sandbox, running tests and reading debug logs — depends on live Salesforce infrastructure for everything except the editing itself. When the platform is unreachable, development grinds to a halt. The author contrasts this with ecosystems such as Java or Go, where compiling and testing locally is routine and an outage at a vendor does not stop work.

As a workaround, the author describes having worked straight through the disruption using a local Apex toolchain called Nimbus that they built, which runs unit tests on the developer's machine, executes SOQL and DML against an embedded database, and simulates platform events offline. Only the final deployment had to wait until Salesforce recovered.

Why it matters

For organisations running sales, support and internal processes on Salesforce, a global login failure is a hard stop on business operations, not merely an inconvenience. The incident is a reminder that a single vendor's availability is a dependency like any other, and teams that build everything on top of it have little recourse when it fails.

For developers, the lesson argued in the dev.to post is to decouple everyday coding work from a live org, so that outages delay deployment rather than all productivity. The question also sharpens as AI takes on more of the coding: if generated Apex can only be verified against a live sandbox, then coding agents stall during outages just like humans do. With Salesforce pitching agentic AI on stage while its own platform wobbled, the gap between the marketing message and operational reality was, for one morning at least, unusually visible.