What is strange is that I've seen nearly the opposite effect with CFL's. The crappy GE's that are far from instant on have been short lived. This was true even when they were turned off/on only about once/day (I don't think I have any of the GE 60W equivalents left for this reason...they all burned out rapidly being used as "nightlights" for the kids.) The instant on types I've used have been long lived. And I've been keeping quite a few in short duration services like garage door openers, bathrooms, closets, etc. I've seen high failure rates with incandescents in this kind of service, but so far I haven't seen that with instant on CFL's.
How many times a day do you turn on your closet lights? The duration is less of an issue than the number of on/off cycles. The classic burnout is bathrooms with short-timer occupancy sensors that might cycle a dozen or more times per day, when other sockets typically see fewer than five.
There have been some bad designs out there that fail from cheap capacitors in the ballasts- your 13W GE twistys might fall into that category, or they ran too much current in the filaments during the on cycles(?). Your experience would not be typical of the industry as a whole. Instant-on cycle number issues aren't "instant death", by any means. I have some instant-on 11W GE globe bulbs in the bathroom that have been going for coming up on 4 years now despite numerous daily cycles. And I have some 14W delayed start twistys from GreenLite that have been in service for more than 15 years in other sockets as well as a handful of programmed start 19W R30s from TCP that are of similar age, with a gazillion hours on them. The adhesive that held the ballast housing to the edison-base gave up on one of the TCPs a year or so ago, but it was still working, despite being visibly crooked and loose (but not dangling by the wires). I had to break the wires and use pliers to unscrew it, but after more than a dozen years service that bulb didn't owe me a thing.
Without data logging it's all about perception (capecod12 seems to think a 100W incandescent has lifespan just 'cuz they get dusty before they burn out
) I'll take the word of those who design this stuff. You'll find that the labeled lifespan of instant-ons are shorter that those that hesitate, and almost all longer-life CFLs have a programmed start delay to be kind to the filaments.
Hoarding 100W bulbs is silly. They're sold even in places with outright bans (like Germany), as space heaters. Given their luminous efficiency, that's at least truth in labeling.
DonL: The incandescent with the diode in series has the same number of on/off cycles, but idles along at a much lower temperature which dramatically reduces the rate of metal being vaporized. From a bulb-life point of view it's no different from using a dimmer running at about half-power. But imposing DC load biases on the grid isn't being very nice to the power transformers. A few 100W bulbs aren't going have much affect, but at higher power half-cycling the DC bias can cause the iron cores of the transformer to saturate, creating a world of current & voltage regulation hurt. A dozen or so years back I designed multi-kilowatt heating element controls maintaining the chemical temperatures on 2-part spray foam equipment that modulated the power output by counting half-cycles,switched on the zero crossing. The firmware layer kept track of how many half-cycles were in each direction, using algorithms for never letting one get ahead of the other, independently of the average power, to keep from smoking the power transformers to the building.