I've verified that the pipeline is working and sending valid records to dev master. I also verified two specific bibcodes from our email conversations:
2022NatAs.tmp....4D has two dois: ["10.1038/s41550-021-01558-y", "10.48550/arXiv.2201.05617"]
2015ApJ...815..133S has one doi, instead of two identical copies of the same doi (so the "list(set(result))" in solr_adapter works as expected).