fasterq-dump is reading something that is not a real run accession
This appears when the thing being dumped is not a valid run: a BioProject ID (PRJNA…), a filename, or leftover partial data cached under one of those names by an earlier failed attempt. If two different inputs give the identical error, it is almost certainly stale cache being reused rather than a real resolution.
Clear what got cached under those names (look in ~/ncbi/public/sra/, or wherever vdb-config points), then loop over real run accessions one at a time — prefetch and fasterq-dump take one accession per call, not a text file of them.