This is not correct - cutadapt supports demultiplexing of paired-end reads. Can you provide a link to this note?
This isn't a particularly huge dataset, and should be able to be demultiplexed with cutadapt pretty easily (I would think). As I mentioned above, I think this might be a problem with your computation environment. Please try demuxing with another tool, or on another computer.
Cutadapt was running an older version (which caused the issue I mentioned previously), but when we updated it again, it appeared to run. However, when we looked at the file sizes for the forward and reverse reads, we noticed they have very different file sizes (297K vs 4.1M). Additionally, here are the first few lines of one of the fastq files:
I just wanted to let you know that we’ve finally solved the issue! We updated qiime2 to version 2018.6, re-uploaded the raw data, and started everything from scratch, and the Bray-Curtis plot now looks normal.