Other gimmicks aside ... there are some definition issues, I believe.

The settling time of the relay, i.e. the time from when it is driven to its active state to when it is in a stabile, fully closed, or fully open, state, has several segments, as Kai has pointed out, yet a part of the time has been overlooked, or, at least, not mentioned. When the contacts are "bouncing," it's easy to assume they'll be travelling between the "on" and "off" states, but I doubt that's the case. I believe that most of the "bounce" time will be consumed in that interval when the contacts are bouncing between the closed and not-quite-closed state, in the open-to-closed transition. While the contacts may, in fact, be in both fully open and fully closed state for a time, that's not likely to become apparent with a single-pole single-throw relay. In any case, there will be interruptions in current flow throughout the bounce interval, and that's what has to be measured.

Clearly, the interval between initial activation (drive) to the input circuit (coil) of the relay up to the first contact, is not being measured in this case, though it is very much of interest. Clearly, the way that Jez has suggested is a solid way to estblish that contact is still intermittent. His suggestion of an integrator is a good one, as it takes into account that there is no current flow part of the time. The end of the bounce time is, of course, when the current flow is maximized. The problem, then, is to detect when the charge rate of the integrator is both maximal and constant. For a normally-open relay, this should yield the bounce time. It's the same process for a normally-closed contact, but it's inverted.

Conclusion: The relay, for discussion's sake, is normally open. There are two time intervals of interest. One is the activation delay, i.e. the time from turn-on to the first contact "make" in a normally-open configuration. Then, there's the interval when the contacts are occasionally open and occasionally closed (the "bounce" time). The final state is when the contacts are closed and remain so in a steady state.

At first, there's no current flow, so the rate at which the integrator is charging is constant (zero). You start charging the integrator when the first contact is made, and wait until the rate of charge is constant. It seems to me that this will apply to both normally open, and normally closed cases. Further, it seems quite unimportant whether it is of a given value, so long as it is constant. There's a "gotcha" in this method, however, and that's that the bounce interval will consist of a number of small intervals during which the rate is essentially constant, and it is up to the integrator to "smooth" these steps into a slope. If you can detect the transition between the resulting slope to a constant level, you're home. The crux, therefore, is to select a time constant which will make it apparent that the charge rate is constant.

I'm curious about one thing, and that's the 0.2 ms value that was introduced in a previous post. I think that the last time I looked at reed relay spec's, the transition time, from open to closed, was on the order of 2 ms. The debounce time that was most frequently used, IIRC, was aobut 20 ms. Have reed relays improved that much in the last decade?

RE