I worked on updating some USGS quads, back in the 90s. It seemed like most of the style guidelines were published internally, long before the Internet, and never made it online.
It's fairly common to show two sets of State Plane Coordinate System grid tics on quadrangles that are near a boundary between two zones. With UTM zones, it isn't such a big issue. The amount of scale distortion within a UTM zone was designed to be less than 1 in 1000. At the edges of the zones, it's 1 in 1000 (at the equator). The distortion increases beyond the boundaries of the zone, but you can usually go about 40 km into an adjacent UTM zone before the distortion becomes a concern. (I do have documentation on that somewhere).
I would imagine, based on the State Plane Coordinate System example, that quadrangles on a UTM zone boundary would have grid tics for both zones. I would be surprised if that continued more than 1 or 2 quadrangles east or west of the zone boundaries.
I'm not entirely sure what you mean by "start at zero in the middle of the map". Each zone would use the easting value measured from its own central meridian, which is set to 500,000 meters. At the zone boundaries, you'd have higher eastings measured from the zone to the west, and lower easting values measured from the zone to the east, but you wouldn't go far enough to get an easting of zero.