I ended up splitting my data into individual vectors and then combining them in R. Yes, it'd be nice to load directly into a "data.frame" but it's not too much bother to do it on the R side.

This whole thing was just because my data was so nicely packaged from the database (via fetchall_arrayref) and it seemed a shame to split it just to stitch it back together.

Thanks. Good to see that it wasn't just some obvious thing I'd missed.