Hi,
Apologies for the delay in this. We've got some code which we've tested this out in, but haven't really come to many conclusions as to if it's helpful. We'll be trying to merge this into the next release soon however. Ultimately, if you have a very multi-reference system, then calculating the projected energy onto multiple determinants can reduce its variance a lot, so in that respect it can speed up the convergence. The convergence of the wavefunction itself wouldn't be affected (unless doing something like Quasi-Newton where the projected energy is used as well as a shift to stabilize the calculation). In stochastic coupled cluster, the projected energy does have a role in the death step for composite clusters, so might change the convergence of the calculations.