Ok, let me try to disentangle this a bit.
I'm not saying that I have seen a problem in which blur is induced by tripod use. I am a bit surprised to hear that several people in this thread can reliably induce it by simply putting their rig on a tripod. Others here have said they have seen it, but only very rarely. What I am saying is:
if it is common enough to be worth manually addressing the condition, then it is solvable in various ways, and should be solved. A solution is not simply a feature that has yet to be added, nor is it an unnecessary expense.
The situation is more like an antilock brake system on a car that prevents the brakes from ever holding the wheels at a full stop. Would such a system prevent the brakes from locking up and causing a skid? Yes. Could a person learn to drive a car that required the driver to disable the antilock system when parked or at a standstill? Sure. But isn't it better that our cars can distinguish between a skid and a standstill? I think so. That's all I'm saying.
When a problem solving mechanism is capable of causing the exact problem that it is intended to solve, and it does so under certain recognizable circumstances, then it should be built to recognize the conditions and to disable itself. A blurry picture isn't nearly as serious as a car accident, but we buy expensive, carefully chosen equipment to increase our success at capturing images, and so if Pentax thinks this is a likely enough failure to warrant alerting us to it, they are effectively saying "we know we need a tripod sensor mechanism, and by the way: you are it".
If that's the route they are going for the moment, then an SR override switch would be acceptable. If I'm the sensor, then let me do my job with minimal effort. A program that recognized the circumstance, and responded automatically in exactly the way we are apparently supposed to respond manually would be better.
vonBaloney is right to point out the presence of an SR override in the Info screen. I had no idea, and it didn't occur to me to look for it there because "Info" is a funny term for that level of control. That placement is not as good as a switch I can feel, but it is
much better than layers deep in a menu. It is easy to overlook until you know it is there, though. I may leave it 'selected' so that a single button-push brings me right to it.
And I really like DeadJohn's point about the 2 second delay acting as a failsafe against failure to re-enable SR. If I'm shooting a critter, that tactic isn't useful, but when it applies it is bound to save some handheld shots that follow tripod use. Until a shooter is used to the routine of turning SR on and off, a person could forget to re-enable SR and soot for a long while before realizing it. Which is exactly why an automated system would be better.
Actually vonBaloney's analysis and metronome video suggest an elegant solution.
Originally posted by vonBaloney So the starting point for SR is always under movement, and so it is always compensating, always creating its own movement, compensating for that, etc. I'm not sure I believe it will get into a positive/amplification feedback loop, but if it is just sitting on the tripod (which usually allow some vibrations when untouched that need to settle) I can see how it might never come to complete rest -- or it would take a good while a least. However if was sitting on something truly unmoving -- the ground, a rock, a solid table -- with no wind, etc it would do better. Seems to me I've actually tested this a bit long ago, but can't remember what I found out. Probably not to worry about it or I'd remember.
I was wondering about the rock vs. tripod comparison myself. If the tripod is problematic because it allows tiny corrections to be amplified, thereby requiring correction, which require correction, which require correction... (a hypothesis for which the rock is a good test) one could program the SR system to alternate phases such that self-induced tripod movement would self-damp. There would be a tiny cost in the moment of phase change, but it might well be undetectable.