A timesheet submission triggers multiple independent downstream steps implemented by different teams; each team will handle its own errors. How should you architect the integration?

The correct approach is to use a single Pub/Sub topic for timesheet submissions because it decouples the submission process from the downstream steps. Each team can then create its own subscription to this topic, allowing them to independently process the timesheet data at their own pace and handle errors without affecting other teams. This asynchronous, fan-out pattern is highly scalable and resilient. Deploying a Cloud Function for each step that calls downstream systems introduces tight coupling and makes error handling and retries more complex across independent steps. Creating a Pub/Sub topic for each step is unnecessary overhead and doesn't leverage the fan-out capability of a single topic for the initial event. A microservice on GKE that calls each step sequentially introduces tight coupling, synchronous processing, and a single point of failure, which is unsuitable for independent, error-tolerant downstream processes.