Certain postscript files made by the freeware plotting program gnuplot experienced a mis-placement of plot elements when displayed by ghostscript, but not on actual PS printers. This only happened when very long paths with sometimes tens of thousands of short rmoveto and rlineto pen movements were used. Gnuplot does that in an effort to minimize the postscript file size. The effect is resolution-dependent, too: turning from 96dpi to 600dpi in GSView makes it disappear... It seems that ghostscript quickly accumulates rounding error in short movements, to the point where the final accumulated error is large enough to be easily spottable on screen. We've since added a somewhat ugly workaround in gnuplot (forcibly interrupting the path and doing a stroke every ~100 steps), but it'd sure be nice to see this quenched at the suspected source.
Ghostscript should be fixed indeed but the work around in gnuplot can be improved. For instance, gnuplot can convert relative commands to absolute commands on PostScript level. *** This bug has been marked as a duplicate of 687173 ***