As a catalog maintainer, I need to understand what the OLM project defines as the best practices for building and maintaining operator catalogs in the file-based catalog world. To build a robust file-based catalog pipeline I would like advice on the following:

1. What should the general structure of my catalog be?
2. Should my catalog follow a push or pull-based model? Put another way, should my catalog repo contain the raw index metadata (operator authors push updates to my repo), or should it contain references to other repos/indexes (operator authors push updates to their index repos/images, this repo pull those updates periodically or on some trigger)?
3. If I want my catalog to support multiple cluster versions, how should I organize version-specific index metadata?
4. Should I provide any automation for operator authors? If so, what should that automation do?
5. Should I build CI/CD into the catalog pipeline? If so, what changes (if any) should require human review?

AC:

• Draft design for OLM catalog maintainer best practices
• Example catalog repository that contains a few sample packages, a dockerfile for building the catalog, and a few PRs with some examples of changes operator authors might make.