With the similar input file (detector volume was different in the old and new input), the relative error is printed as zero in the sum file when I used the new fluka version. I tested some other input files. For some, I observed similar zero relative error printed. The dominating energy bin is the last bin in the tab.lis file, where the error is printed as zero, hence resulting a zero error in the sum file.
This does not seem to be related to the new FLUKA version, where no relevant modification was introduced, rather to the change in your input file. Was the three orders of magnitude (!) increase in the detector volume corresponding to a consistent modification of the geometry region? If so, it’s not surprising that the statistical error considerably dropped, down to the threshold (1e-6 of the scored value) below which it’s considered as zero.