SLooking at the Z Yoke design, which is about the same for printed and CNC metal, there is certainly room for improvement. 

The linear rail car is constrained from any rotation and linear translation in the X & Y, free to move up & down on the Z. The lead screw provides Z positioning. Below shows the linear constraints (red) rotational constraints (blue) and the primary force path from the lead screw nut to the bed holding interface.

The challenges here

1. Yoke to linear rail carriage is only positionally constrained by the screws. 
2. Lifting force from lead screw is loading the lead screw nuts in tension, loose screws will affect the height
3. Method of bed mount is sub-optimal, not really allowing for misalignment and will have binding issues unless there is excessive clearance, which could cause bed movement

My plan:

• Improve the alignment between the yoke and the rail carriage
• Mount the lead screw nut from underneath, so that weight on yoke is resting on the nut
• Add a simple kinematic mount for the bed
• Shift the mass around so it's working to stiffen up the yoke 

This is the direction I'm heading, although this will take a bit longer to get to where I want it to be.

• Kinematic mount for the bed, different parts for each of the 3 yokes
  • Left lower corner will capture the ball, point constraint 
  • Left upper corner will capture the ball on 2 pins (resists rotation on X & Z)
  • Right side will provide a flat plane (resists rotation on Y)
• Kinematic mount will use 8mm Si3N4 ceramic balls, which has low thermal conductivity
• I'm unsure if the lead screw nut has clearance holes or they are tapped, so need to determine that.
• Still determining the spring to use here

Concerns I have at the moment

• If I'm going for a heated chamber, the linear rails are not typically rated for above 80°C
• How committed am I to trying to stay with the RailCore design? Will I allow myself to deviate substantially so that I get the design I want without being backwards compatible?
  • I'm starting to see things that I don't really like in the RailCore design, but fixing it will stray me far off the path. Or do I look at a clean slate design here? Going to sleep on that one.