These authors find reasonably secure counterparts for 35/38 SCUBA sub-mm sources in GOODS-N. This plot shows the relationship between redshift (combination of spectroscopic and photometric redshifts) and rest-frame 850 micron luminosity (which is also illustrated by the size of the circular plotting symbols). The black symbols are those below the median luminosity, and the blue symbols are above. The curves show the detection limits; since the noise varies strongly over the SCUBA map, a different completeness curve is shown for the location on the map of each detected galaxy. The fact that these curves flatten at high redshift is due to the negative K-correction.
One notes an apparent lack of high-luminosity sources at z<~1.5 and of low-luminosity sources at z>~2.5. This is interpreted as an aspect of the generic downsizing trend of star-formation. The authors split the sample in half by luminosity, and model the evolution of the luminosity function of the two samples independently as a function of redshift. But the qualitative conclusions remain the same, and based on the data you'd have to be pretty brave to conclude very much more than that. I'll post a few more thoughts in the comments, but those will probably only be relevant to those who have actually taken a look at this paper.