https://forum.qiime2.org/t/problems-with-demultiplexing-dual-indexed-pair-end-sequences-from-miseq-illumina/21571
This seems to contain the answer: Victoria used two metadata files. I used the same column for the forward and reverse reads, as they are the same. Making my metadata file like this:
seemed to do the trick. Atr least it is still processing...