Salesforce lost much of its core platform for most of Wednesday 16 September, the second day of Dreamforce. The Register’s running account has hundreds of instances affected across the US, Europe, India and Japan from about 07:50 UTC, with the incident declared resolved around 19:20 UTC. Salesforce traced it to requests stalling on an internal login service and exhausting server resources.
For email teams the interesting detail is what stayed up. Incident 20004433 was filed against Core services. Marketing Cloud was listed as available throughout, according to one consultancy’s reading of the status page.
That follows from the architecture. Marketing Cloud Engagement is the former ExactTarget, bought in 2013 and still run on separate infrastructure. Salesforce’s Trailhead material lists this as a constraint: legacy architecture, outside modern Core. On Wednesday the constraint was a firebreak.
Marketing Cloud Next is the platform Salesforce wants marketers on. It is built natively on Core, with Data 360 as the data layer and Agentforce throughout. Segmentation, orchestration and sending share a platform with Sales and Service Cloud. When Core stalls on a login dependency, a marketing product built on Core has nowhere else to stand.
We have not confirmed that Marketing Cloud Next customers lost sending on Wednesday, and have asked Salesforce. The exposure is a matter of design either way.
Engagement customers should not feel smug. The sender stayed up. Much of what feeds it lives in Core. Marketing Cloud Connect pulls CRM data into Engagement and fires journeys from CRM events. With Core unreachable, those triggers had nothing to fire from. Flow email alerts, case replies and Experience Cloud password resets are sent from Core itself. Salesforce also said it had reports of “scheduled jobs not running as expected” on instances that had come back.
Recovery carries its own risk. Jobs that failed get retried, syncs catch up, and queued triggers release at once. For a sender that means volume spikes and the chance of duplicates. Consent is the sharper concern. Where opt-out status is mastered in CRM, a ten-hour gap is a window in which an unsubscribe may not have reached the system that decides who gets mailed next.
Three checks for Salesforce senders this week. Compare Wednesday’s triggered volume with a normal Wednesday. Look for duplicate sends on Wednesday evening and Thursday. Reconcile unsubscribes captured during the window against CRM consent fields.
The longer question is for anyone planning a move to Marketing Cloud Next. Unified data is the pitch. A shared failure domain is the price. It belongs in the evaluation.
