I’m building a file-processing service that may receive uploads from Amazon S3, Azure Blob Storage, and Google Cloud Storage. The event mechanisms differ, but I’m trying to keep one processing pipeline across all three.

How do you design around duplicate object notifications or out-of-order storage events? I’m especially interested in what you use as a stable event identity, and whether you rely on object versioning, a processing ledger, or another approach to prevent the same file from being handled twice.