fix collection hid inconsistency

reported by @Delphine-L

If you run a tool that creates a collection it gets `hid: 1`. After the contents of the collections are created (let's say it contains 3 datasets) it will be re-hided to `hid: 4`. If you schedule a job with that collection as an input before it gets re-hided you will get a datasets named `job X on hid:1` - but hid:1 is now a hidden dataset - a member of the collection with hid:4.

The issue names no files, tests, or entry points. Start by reproducing the sequence of creating a collection, adding three datasets, and scheduling a job before re-hiding completes; trace how the collection and member identifiers change. Done means the scheduled job continues to reference the intended collection or dataset without an identifier collision.

Written by the indexing model from the issue text.