Belt tensioners on toothed belts

I just finished helping a customer with a servo driven plasma table
and I was surprised to see spring loaded idler wheels on swing arms
(like the old reel-to-reel tape drives) being used to tension the
belts on the toothed pulleys. The tension mechanism was a coil spring
and was not very taunt so that when the motors changed direction the
belt would bow out or in, depending on the direction. Seems to me the
flexing of the belt would introduce some errors. Do any of you have
any experience with a design of this type? Seems like a way to move
the motor to maintain belt tension would be a better approach. I
guess if the belt tension was strong enough it might be a better
solution (?)

I do know that in this circumstance it ended in disaster. The pivot
arm became loose and the arm got cocked at an angle. It was okay with
the axis running in one direction but it acted almost as a brake in
the other direction and since the motors were large it just ran until
the fuse blew (about 30 msec after the servo drive blew!).

Tom Caudle