Shaving down the TO-92 package should reduce its thermal resistance ... but that still doesn't manage the >60V voltage?excursions that are usually the killer of BS170s.? Those are generally just user error.?
If > 60V excursions are the usual killer of BS170s, do you have an actual defined load circuit that will by measurement, produce over 60V peaks at the drains of the BS170s? If so I'd love to know, because as yet I have not had time to rig up some tests. It would be very interesting to have an actual case to replicate and test.?