The fringe tracker in MIA+EWS tries to zero the OPD and (absolute) phase in every measurement.. The absolute phase is not known but water vapour adds a certain amount in every wavelength bin. As the amount of water vapour changes, the phase change varies as well. If the phase is not removed from the measurements, subsequent frames could no longer be added coherently. In previous versions this was done in such a way that all phases were biased towards zero by about 1° and this subsequently let to too large correlated fluxes since F_{corr} = visamp \cdot \cos{visphi}. In the new version this water vapour correction is done in a different, unbiased, way. This is only relevant for weak sources (F_{corr} < 0.5 Jy). On NGC 4151 data the new procedure leads to a correlated flux that is about 0.85x the flux from an older EWS version. On a bright source the difference between the old and the new version is negligible.
The new function meddelay (not in the official release yet) runs a median on all unflagged data points. This reduces the correlation effect between individual data points. There is not a big difference between the correlated flux with / without this function from what I have seen.