In DM-29266 code was added to translator classes to teach them to read multiple relevant headers from a single file.

This ability needs to be added to write-sidecar and write-index to allow these files to be used to ingest multiple dataIds from a single data file efficiently without requiring a file download.

Some thought is required as to how to arrange the content.

• Sidecar files could in theory use the same schema as index files but would the key names be integers? Would a list of dict work better?
• Index files could treat everything as completely different "files" but using file#1, file#2 as the keys. Or else the content could match whatever is chosen for sidecars.