The usual CNC plasma table is simply made of rectangukar tubing fabrications tweakable to surprizing accuracy. If the rack is dialed into the axis motion, backlash once set will be fairly consistant. Backlash is usually compensated by a setting in the CNC control and if a little scientifically applied axis friction was part of the design backlash will play no significant role in the table's tracking and positioning accuracy..
Rack and pinion loads in a CNC plasma cutter are mostly incidental: inertia, out of level, accumulated dirt, and wire crag are probably the main culprits.
If backlash is a problem, simple spring loaded mounts holding the axis drive motor that act perpendicular to the pitch plane will tight mesh the rack and pinion ensuring zero backlash. The compliance of the spring loading ensures relief should it be necessary thereby avoiding mesh interferance and possible axis jamming..
If you're concerned about available rack lenghts being too short for the desired motion , don't be. Rack can be butted precisely by meshing and clamping a short length of rack remnant to the lengths being installed. If the butt ends are fitted with clearance at the tooth space root then the rack remnent spanning the butt will place both rack incorrect pitch relationship. Then it's just a matter of dial, clamp, drill and tap to attach and align the rack