We identified an issue where custom domains deployed for certain clients triggered a configuration conflict with the Clerk satellite domain within shared workers.
This doesn't affect most online users, impacting only a specific subset of clients. The root cause is that this domain-related configuration prevented Clerk from routing authentication requests to the correct API domain, causing those auth requests to fail.
This is an incident related to a specific production environment. We're currently thinking about how to automate it, and how to add guardrails for cases like this.