There are two proficiency testing studies that were done several years ago on Proteomics. One by HUPO and one by ABRF. These did not specifically address the reproducibility of trypsin digestion, but they are telling in what they revealed about lab-to-lab reproducibility of proteomic methods, which are pretty awful.
We test our trypsin with the standard assays (listed by Sigma-Aldrich) prior to use. About 1 in 3 lots have way too much other generic protease activity and the enzyme activity of each lot is highly variable, so you need to standardize on the activity level of your particular lot to the protein content of your sample to get any reproduction in the method. It only gets worse with other endoproteases.
HUPO Study: Bell, AW etal. Nature Methods 6(6):423 (2009).
ABRF Study: Friedman, DB etal., Proteomics 11:1-11 (2011).