Media handling in content pipelines

Video files are not the same as text data. They're large, processing takes time, and errors in the middle of the pipeline are more costly.

We use object storage for all media files — this solves the scaling problem but creates new questions: how to manage file lifecycle, how to avoid storing unnecessary data, how to serve content quickly when needed.

One thing we figured out: file upload and file publication are different events, and there must be a clear state between them. A file that is uploaded but not yet published — that's a valid state that needs to be handled correctly.

Details matter.