I’m investigating.
Usually when topojson appears to hang, it’s because it’s effectively out of memory and it’s spending all its time paging. TopoJSON 1.4.0 introduced a new, more memory-efficient hashtable; however, it’s possible that the linear probing in this hashtable implementation is causing frequent collisions, which would make the code slower.
So, I’m investigating to see whether there’s a bug, or a potential optimization, or if more memory is needed to convert such a large input file…
Of course, since the zip code tabulation areas are so large, you rarely want to create a TopoJSON file from it that you would serve on the web, so you might have more luck either slicing it into tiles (raster or vector) or rasterizing the whole thing as an image by reading the shapefile [1] directly.
Mike