Closes #16573.

When a collection using @payloadcms/plugin-cloud-storage has a custom afterChange hook that throws an error, the plugin swallows the error, logging it as a warn. Payload silently rolls back the changes, leaving S3 and the database in inconsistent states without notifying the user that an error has occurred.

Problem

Two problems in the plugin's afterChange handler:

1. The inner req.payload.update() was wrapped in a try/catch that only emitted logger.warn. Anything thrown downstream — including a custom afterChange hook — was swallowed, so the outer operation completed "successfully" while its side effects were broken.
2. The old files were deleted from cloud storage before the new file was uploaded. If anything threw mid-flight, e.g. from a user-defined custom hook, the old files were already gone and there was no recovery path.

Solution

• Remove the swallowing try/catch around the inner update. The outer try/catch in the same hook already logs and re-throws, so errors now surface to the caller and the admin UI.
• Wrap the skipCloudStorage context flag in try/finally so it is always cleared, even when the inner update throws.
• Reorder operations: upload new files → persist adapter metadata → delete previous files. The deletion step now also skips any filename that matches the new doc's filename, so reuploads that reuse a key don't delete the freshly uploaded object.

Now, custom hook errors are surfaced to the caller, the previous file stays in S3, and the document keeps pointing at a valid object.

• To see the specific tasks where the Asana app for GitHub is being used, see below:
  • https://app.asana.com/0/0/1214708232205239