I just pushed 0.1.24, a total rewrite of step3 preview mode and refmap unmapped reads handling. Should be much faster, and preview now actually previews all data in step3, rather than just the reference mapping part.
I tweaking preview mode in step2. Currently when it selects only the first chunk this makes it so that only one processor runs on the data. Instead I'm going to slice a chunk of data out and let zcat_make_tmps still split it into bits to be spread across multiple processors.
Maybe I can generalize preview_truncate_fq to work for both steps 1&2 in this way.