Probably easy to fix (just update the tests data) but right now I can't look into it.
We could maybe as a long term solution use some mock framework to replicate the current NCBI expected behaviour and not rely and any specific metadata.