There is no question that exploding dimension is a bad idea. But if the only other choice is paperspace dimensions that later change their values WITHOUT user intervention, exploding is a better choice.
I think the problem is uncommon. But, if only 1% of companies have this problem, and they catch the AutoCAD error 90% of the time, that would still mean 1 out of 1,000 drawings might have a dimension that spontaniously forgets its scale factor. Is that an acceptable risk?