Thank you for this contribution. In principle this looks quite good, but I have one issue. Although these records pass the validator they have some problems. In the "COMMENT" section i can find information like "COMMENT: Origin: Animal, CSID: 10334, SubCategory_DNP : Lipids, CASID (tmp): [18951-77-4], Fatty acids". Better would be to place things like "CSID: 10334" and "CASID (tmp): [18951-77-4]" in the CH$LINK section. Please check CH$LINK
I would rather see it in this way:

CH$LINK: PUBCHEM CID:10334
CH$LINK: CAS 18951-77-4

in the recordfile.

Do you think it would be possible to create the records in this way? Do you use your own software to create the record files?

Nice work. Thank you for your contribution.

There's still problems here. Things listed as CH$LINK: PUBCHEM CID should really be CH$LINK: CHEMSPIDER instead! At least for the ones I checked.

Yes @meowcat, you are right. I tested some and it was never correct. Automatic validation doesn't check this yet. Implementing test for this is on the roadmap, but I'm already a bit afraid of the numbers of mistakes I need to fix...

I think it will be hard to validate this strictly, since there are multiple true and half-true answers sometimes (stereoisomers, mixtures, salts etc will all not have a simple answer).

It should be easy to check, the ChemSpider ID and PubChem CID should be an InChIKey match, at the very least an InChIKey first block match. Everything else is clearly wrong. Entries that fail an InChIKey check should be validated.
Could this case be a misassigned identifier?