I recently read an article about the 5 common mistakes when selecting linear guides... one of them was related to selecting linear rails/carriages with tighter internal tolerances than the mounting surface/points.

So I have a general application question about 3D printers and the best approach to linear rail selection... Oftentimes, many of our printers are made up of an assembly of aluminum extrusions to form a cube or similar shape. On one of the axis (usually Y), some people use a parallel linear rail system which carries the X gantry for the print head. Since the two parallel linear rails are mounted to separate extrusions that are connected using perpendicular extrusions and corner gussets, the parallelism of the mounting surface certainly won't meet the specs outlined for most linear rails in parallel. Obviously some sort of adjustment must be made to try and make them as parallel as possible. What is the best approach for this sort of adjustment on typical t-slot style extrusion (ie. 8020)? I usually don't see reference shoulders or pins, or fine adjustment screws... it looks like most people just mount them directly to the extrusion slot with t-nuts or something? Not sure how they are ensuring any sort of parallelism?

Based on the article I read it would seem that it would be recommended to select linear rails with some clearance in this very not precise mounting application to prevent binding and unfavorable wear conditions... However, if there is too much clearance it seems possible that some amount of "slop" will show artifacts in the printed products due to the play in the bearings. Would it be advised to externally preload the X gantry by introducing a small but constant force on the X gantry perpendicular to the Y axis linear rail to pull the ganrty to one side (of the Y axis) and take up that play? Or introduce a small moment on the gantry to take up the play in the Y axis? Would doing this end up causing more bearing wear? Or would it be more ideal to remove the parallel rails altogether and use a larger single rail with bearing carriages farther apart (to handle the moments on a single rail and eliminate the parallelism problem) and the opposite side of the gantry just have it rolling on flat surface unconstrained? Would love to know your thoughts. Thank you!