One view over several stores
The state of a single file was spread across multiple DynamoDB tables and MySQL databases, because the services that touch it each own their own. Aggregating that into one view is the whole product: the data existed, it just could not be read as one answer.
Recovery, not just observation
A monitoring tool that only reports failures moves the work rather than removing it. The recovery mechanism retries failed AWS Step Functions automatically, which is what turns this from a dashboard into a service that reduces manual intervention.
Serverless where it fits, containers where it does not
The backend is AWS Lambda deployed through AWS SAM: the workload is event-shaped and idle most of the time, which is the case serverless is actually for. The UI is a long-running service and sits on ECS. Both ship through end-to-end Jenkins CI/CD.