Partialator deltaCChalf seems to result in twinning
I've had a report of apparently twinned data from partialator when using large numbers of iterations. The fact that it only happens with larger numbers of iterations already narrows it down a bit: with no post-refinement, it should be almost completely stable regardless of the number of iterations. On testing, adding --no-deltacchalf
removed the apparent instability.
I don't know if this is expected for the algorithm, or due to a bug. In any case, the problem has been there for a long time...