Currently the correlation scatterplot just takes the two trait/dataset names as input (so the page can be linked to others). The problem with this is that it means that any edits to sample data a user does for correlations are not reflected in the scatterplots.
I'm thinking of dealing with this similarly to the way I dealt with the mapping links - just using the hash as a key in Redis that is linked to the sample values, and having the key be used as in the URL instead of trait/dataset (when values have been edited).