Sounds like an overflow issue. However, iso-volume may not be the most memory (or computation) efficient here since it will generate an unstructured grid for the full extracted volume. If you’re just interested in the low-iso-value high-iso-value surfaces, contour filter may be the better option.
Another option is to run with a parallel pvserver, so that the data gets splits into multiple chucks and thus avoid the overflow.
I am going to try to see if I can determine where the overflow is happening locally.