If you take a look at SuperLatticeTimeAveraged3D.hh, in line 64 (current Doxygen), you will see that the expression matches the expression I wrote in my first bug report message.
Mathematically, the cross correlation of two fields ends up being equal to
CrossCorr = (uv)_av – u_av v_av ,
where _av denotes a time average. Each time average is the sum of N observations divided by _ensembles.
Thus, the first term requires a single division by _ensembles and the second term requires a double division by _ensembles.
In my simulations I notices that the Reynolds stresses never converged and kept increasing in value.