Until now, lacking hard data, I was a believer that instant start is worse for T8 lamps than magnetic rapid start and shaved away more life each time the lamp is started.  It's quite the opposite and by a significant margin based on a study.  

Industry standard testing is 3hrs on and 20m off, but obviously to experimentally obtain data naturally, it would take way too long, so to study the effect of frequent switching, they used a 5m on/5m off cycle.  

They tested seven different models of ballasts and each model of ballast was given a sample of 36 lamps.  All the ballasts tested were 2 lamp type, so I think they had 18 units of each ballast model going at once, each driving two lamps.  

The median cyclical lamp life of two magnetic rapid start ballasts were 2,800 and 3,100 cycles.

The median cyclical life of two instant start electronic ballasts were 13,500 & 19,500
Values were similar for two electronic rapid start, although one electronic rapid start hit 43,300.

This shows that in short cycling, which is often seen in residential use, lamps last five times as many switch cycles compared to rapid start magnetic.

So, I conclude that in frequently switched F32T8 applications, the lamp life from best to worst is:
A. Well designed electronic programmed rapid start
B. Electronic instant and rapid start
C. Magnetic rapid start

It's also well understood that magnetic gives the worst system efficacy.  

You can see the report here:
http://www.lrc.rpi.edu/programs/NLPIP/pdf/VIEW/Guide3.pdf

I think the reason is, then most so called "rapid start" ballasts have so high OCV, so they in fact start the lamp instantly.
The reason is, then for true "rapid start" the OCV should be only just enough to ignite the lamp with properly heated electrodes, but not enough to ignite lamp with electrodes not yet reaching this temperature.
But as this voltage is subject to large lamp-to-lamp variation (it is influenced by exact electrode shape and position) and variation given by by operating conditions (grounded shield distance, surface conductivity, temperature, lamp age,...), so there is no OCV matching for all applications. These variations are even way larger with thinner lamps, that's why so many issues with T8's in RS.
As the too low OCV is considered worse (lamp does not start), usually it is designed rather high (to ensure ignition at cold and with humidity), so in most application the arc ignite too early, yielding cathode wear.
On top of this "high efficiency" RS ballasts use capacitors as main ballasting impedance (it has nearly no losses, compare to inductors), what yield to quite high crest factor due to lamp nonlinear behavior, mainly upon start - and consequent electrode wear.

I think regular preheat with glow-bottle would be only a bit better, as even there the starter is usually closed way too short time to preheat electrodes well (most of the turn-ON delay is the starter itself heating up to close, but during this time lamp electrodes are not yet heated).

Plus since the magnetic ones use a lower OCV than the electronic ones, it's entirely possible that the lamps that stopped working on the magnetic ballast might still work on the electronic ones since they will drive it with a higher voltage. Plus I don't think I've ever seen a lamp removed from a magnetic RS ballasted fixture that had a big black spot and a crack or hole at one end before either.

That's kinda weird...for programmed start...DEFINITE AGREED! But in my store....our backroom bathrooms.....and front room offices has instant start and controlled by motion sensors, they only last 2 months in average!

The programmed start in back room also using motion sensors has been working 5 years though

I've seen rapid start at homes that has been running for 30 years!!!!!

My thinking is the NEWER magnetic rapid start don't last as long as the OLDER magnetic rapid start....

On top of this "high efficiency" RS ballasts use capacitors as main ballasting impedance (it has nearly no losses, compare to inductors), what yield to quite high crest factor due to lamp nonlinear behavior, mainly upon start - and consequent electrode wear.

Did you read the PDF? It's stated that all the ballasts met ANSI specs.  

Here are ANSI specs for reference:
http://www.lrc.rpi.edu/programs/NLPIP/pdf/VIEW/Guide1.pdf

If it was predominantly capacitive, I doubt it would meet <1.7 LCCF ANSI specs.  They didn't disclose the make and model of ballasts tested, but magnetic F32T8 ballasts are quite heavy.  Since NLPIP don't accept funding from manufacturers, I think they're quite bias free.  

I think two different commercially available models of F32T8 magnetic rapid start ballasts and a lamp sample size of 36 for each type, the data is quite reliable.  

The OCV for the magnetic ballasts tested were 325v, LCCF 1.34 and 350v, LCCF 1.5.  

If you look at the graph, you'll see that on black filled circle ballast chart, over 50% have failed by 3,000 cycles, but a handful surviving after 25,000 cycles under same setup, so I don't think an anecdotal evidence of a fixture or two lasting a long time is statistically significant, no more so than saying "I've heard of a man living to 120, so I will too".  

This experiment is for F32T8, which have drastically different operating voltage vs starting voltage profile than F40T12s though.  So, what maybe true for F40T12 may not be true for F32T8. 

CCF 1.7 is rather high, but i thing still achievable with HX autotransformer + capacitor (normal assembly of HPF, low loss RS magnetic ballast). Pure (EU series choke) inductive ballasts have about 1.4..1.5, HF electronic 1.4.
But these published numbers describe steady-state operation, not the startup.

And 350V might be enough to start warmer 4ft tube without electrode heating, mainly with good grounded shield in the very close proximity.

CCF 1.7 is rather high, but i thing still achievable with HX autotransformer + capacitor (normal assembly of HPF, low loss RS magnetic ballast). Pure (EU series choke) inductive ballasts have about 1.4..1.5, HF electronic 1.4.
But these published numbers describe steady-state operation, not the startup.

And 350V might be enough to start warmer 4ft tube without electrode heating, mainly with good grounded shield in the very close proximity.

Not in the PDF, but in the actual report it cited.

"The OCV for the magnetic ballasts tested were 325v, LCCF 1.34 and 350v, LCCF 1.5.  "
So, we've ruled out LCCF.