To do some testing, i redesigned my model in 3dsmax.
I may have found the problem, and it could have been the box collsion models having more than 12 polygons or 6 faces, i have a habit of dividing my models down the center on all axis so i have center vertices which makes it easy to align centers of objects together.
I may have forgot to remove these extra faces leaving the model with to many face's / triangles.
Interesting thing, 3drad did give me an error "couldn't load rigidbody_.x" but, in 3drads true form it did load the visual mesh part anyway and even allows the project to run with a defective model......
So, if this problem doesn't come back, i'm going to blame this as 50/50, cause both 3drad and i made mistakes here, assuming the problem doesn't return.....
50/50? more like 0/100 [rad/you], in an ideal world rad would've pointed out exactly what the issue was, but any error is a hint and a red light to stop work and investigate the cause until it stops [like view that _.x in meshview or swap it for another _.x] and not continue as if nothing happened?
why rad let's you keep running it after the error is for troubleshooting - by a process of elimination: delete anything associated with the error until you find which one is causing it. Simply stopping you from running it after an error as you suggest would make this MUCH harder, requiring advanced tools like process monitor that i have mentioned before.
and 2nd, if you choose to make custom collision models (i assume is what you did) you are bypassing rads built in automatic sphere group creation and saying 'no thanks, i can do it myself with boxes, etc' [which i also posted about] so any and all error checking is ENTIRELY up to you, and not rad.
for me, i always check exports in meshview to make sure it is ok, before i even THINK about importing it to rad, if you had done this there would be no issue. i hope you learn from this, as i have also learned some things from you.