I believe that soft starting motors will reduce rotor bar problems.  Anytime you can reduce the magnetic forces produced by the large starting currents, you're doing your motor right.  As far as detecting rotor bar cracks and/or perocity, an analyser made by PDMA Corp., Tampa FL. does an excellent job without having to dismantle the motor.  This test is called the RIC (rotor influence check) and it also can detect other problems such as air gap and winding issues.  I personnally used this technology with a great deal of sucess in the petroleum industry.  It is a software based technology that is somewhat spendy ($35-45k), however some motor service shops do provide this testing as a contracted service

6/30/00
Also, the torque behavior will influence the rotor bar life cycle. Any abrupt torque changes shall be mitigated by a flywheel, electromagnetic shaft coupling, etc.

I am not so sure one can fairly characterize cracked rotor bars in a large industrial motor after 20 years service as "severe" bar cracking"  

The fact that a large majority of bar cracking was detected at same time, after decades of service, suggests a high cycle fatigue failure which is to be expected in this type of machinery.

Rotor bar cracking is fundamentally a thermal stress fatigue problem arising from differential thermal expansion of the bar relative to the rotor.  The bar experiences large currents and therefore heating, during periods of high slip, including startup transients.  This thermal stress is to some extent unavoidable.

Assuming a well designed machine, rotor bar cracking is minimized by reducing the number of starts or starting under reduced voltage conditions.

What about a purpose of squirrel-cage induction motors with double cage construction? Such motors, if applicable, have less chance to have cracked cages. Of course, they cost more. One may pay now or later according to one's prerogative.