Accuracies needed by IOTA and others for occultation timings:
1. Lunar grazing occultations have the most relaxed requirements; visual timings of them to an accuracy of +/-0.5 second are fine. The accuracy of analysis of grazes comes mainly from knowledge of the observer's geographical location, where the +/-0.2" to 0.3" in longitude and latitude that can now be obtained with simple GPS receivers (and can still be obtained by carefully measuring large-scale topographic maps) is good enough.
2. Lunar total occultations should be timed to +/-0.3 second or better, to be useful for improving the lunar profile. This is near the limit of visual timing accuracy, so video timings are becoming preferred, mainly to guarantee times to the 0.1 second level.
3. Asteroidal occultations should be timed preferably to +/-2% of the predicted central duration. For a 10-second occultation (most that we predict these days are shorter), that's +/-0.2 second, which is at the limit of carefully-made visual timings. Slightly worse accuracy is acceptable for observers near the limits (somewhat for the same reason that timing requirements are relaxed for lunar grazes), but one rarely knows in advance if this will be the case; in most cases when an occultation occurs, the observer will not be near a limit. So for asteroidal occultations, video timings are really preferred; it is for them that the investment for developing a video capability is most valuable (but remember that lunar events are more frequent). Most asteroidal occultations involve stars fainter than those usually observed during lunar events, so the more sensitive cameras, like the PC-164C and Watec 902H, are important.
Keeping these in mind, when you describe visual timings, considering their accuracy, it is not necessary to correct for time signal propagation time; that is only a few hundredths of a second at most, negligible relative to the +/-0.2 second or so (some can get +/-0.1 second with careful practice) of visual timing accuracies. Also, for visual timings, tape recording is preferred, especially for grazes and asteroidal occultations, over just using stop watches or the "eye and ear" method.
For video, the observer shouldn't need to apply the correction for time signal propagation to his site. This is virtually negligible for lunar occultation events, considering the time errors of a few tenths of a second caused by lunar profile uncertainties. But when we receive reports of timings made with video, we will assume that the light-time propagation has not been taken into account, unless the report states otherwise. If the observer reports which time signal he or she used (that SHOULD be reported), then we can calculate the great-circle distance and apply the correction to good-enough accuracy (of course, the millisecond or so variation due to ionospheric variations is much too small to be of any concern).
For asteroidal occultations, I haven't been concerned with light time propagation, partly because most of these are still timed visually where the errors are larger. Even when timed accurately, they are usually observed from a limited geographical area where the time signal propagation differs by only a few milliseconds, smaller than video frames, for all of the observers.
I would much prefer the largest number of observers using video to time occultations to an accuracy of 0.1 second guaranteed, and often to 0.05 second, than to have only a few who have the electronic background to achieve absolute timing accuracies of +/-half a video frame.
In the future, smaller timing accuracies may become more important. But by the time that issues like the difference between half a video frame and a full video frame become important, the timings themselves may no longer be needed (such as from lunar orbiter missions that will map the Moon in much more detail than has been done so far, and radar or space missions that observe the detailed shapes of asteroids, Gaia for producing a star catalog that will be much more accurate than Hipparcos for millions of stars, etc.).
Certainly for many purposes, less accuracy in the position is all right. For example, for asteroidal occultations, 200ft. would be all right, since that's about the size of diffraction effects at their distance. And for lunar occultations, the largest uncertainty is the profile data, which have individual errors of the order of 1000 ft. However, especially for grazing occultations, the resolution that can be achieved is of the order of 20 feet or so; observers 50 feet apart have observed noticeably different sequences of events (small differences, but nevertheless differences). I think 50 feet is a good value for grazes, and other lunar occultations, too, since at the Moon 50 feet subtends 0.01", which is about the accuracy of Hipparcos star positions. So in order to effectively eliminate geographical position error as a consideration in reductions of the observations, we would like to have the positions known to 50 feet or better accuracy. Such accuracy can now be achieved readily with GPS measurements (a few minutes averaging is enough), and it can also be achieved by careful scaling of a large-scale topographic map.
David Dunham February 2002