Well, here's the conclusion I reached. I've used a rigid probe for most of the last year. But, Mach3, flaky thing that it is, *will* periodically screw up, and break the probe. I've been through at least a half-dozen in the last year. So, a probe with some "give" is, to me, a requirement. But, the typical Renishaw-style probes have several drawbacks, when used for machine setup: First, they have some travel after contact, but before switch actuation, and this travel is different at different angular positions. Careful calibration is required to minimize the effects of this. Also, they are very delicate, and difficult to calibrate.
I have a design for a contact probe with "give". The probe itself will consist of a slender stylus with a 5mm ball bearing bonded to the end. This will give a very precisely spherical tip. The stylus will be supported by a spherical Delrin bushing about an inch above the ball tip. The top end of the stylus will be supported by three springs, each with a tension adjusting screw. By adjusting the spring tensions, the ball tip can be precisely centered to the centerline of the spindle. The "switch contact" is simply the electrical contact of the ball tip to the work, so there is no "overtravel" - contact with the workpiece activates the Probe input directly. If pushed beyond the contact point, the springs and spherical bushings allow the ball and stylus to swivel out of the way, so there's no damage. Once the tip is released, the springs should push it back into perfect alignment.
This has no sensitive mechanical or electrical design aspects, so should be easy to build. The *only* potential downside I can see is "stiction" in the spherical bushing can prevent the tip from self-centering. But, with a little lubrication, and adequate spring pressure, this should not be a major problem.
The other option I've considered is simply making an insulated holder for a "wiggler" edge finder, but that would have to be manually centered after each contact, which would be very inconvenient.
Regards,
Ray L.