After a few days of much-needed vacation, I spent travel time towards Ringberg Castle working out and writing down an estimator for the correlation function and its gradient. That is, it is a replacement of the standard estimator that gives you the correlation function in radius bins with a new estimator that gives you the parameters of a parameterized model of that correlation function, even if those parameters have correlated (or covariant) estimates under the data. This was a straightforward merging of the argument underlying the Landy & Szalay estimator with the argument underlying linear least-square fitting.
I wrote up a telegraphic latex document and sent it to Kate Storey-Fisher (NYU), who might try to implement it later this summer. I'm excited about cosmology again!