MBS-10404 (or maybe an earlier change? anyway) limited the amount of recording relationships returned through the API so that very large releases with over 500 recordings do not return recording-level-rels (nor work-level-rels). This makes sense, but we should have an alternative way to get the missing relationships with more calls. We could have an endpoint that only returns recording IDs + recording and work level rels for a release, paginated (not sure what would be a good number per page, 100? 250? 500?). This would allow Picard and other data users to eventually get all the data, just slowly (but not as slowly as having to query for the recordings one by one).