Bulb warming should not be an issue with cold light, yet this effect is still there. I think it has more to do with hitting the threshold of the paper and then building up density from there. In any case, the effect is real so we just have to live with it.
Paul Strand used to do the same thing Corfield did. He said that on windy days, natural objects would always return to their resting state once the wind died down, so he would just cover the lens when the wind blew and then uncover it again after everything settled down. With all the reciprocity issues at play on long exposures, though, the intermittency effect if probably "overruled" by reciprocity issues in terms of how it affects exposure.
I've long found it to be MORE of an issue with cold light. Depending on the phosphors, a warm cold light (sure and you know what I mean) can give as much as a stop more than a cold one. That's quite apart from start-up times. Or were you talking about using a shutter?
2 seconds to 10 seconds is a log range of 0,7, so the inertia should be overcome in the first 2 second exposure if 10 seconds gives a maximum black, assuming an ISO(R) of 90-110 (medium grade paper). Latent image regression should not, I think, have much effect in the next 4 exposures.
In 1927, Clarence Weinland, J. Opt. Soc. Am. 15:337 'The Intermittency Effect in Photographic Exposure' demonstrated that previous exposure to bring the latent image up to the beginning of the straight line portion of the curve caused almost complete elimination of the intermittency effect (Haist page 747).
Intriguingly, Haist also suggests (on page 746, citing Silberstein and Webb in 1934) that 'when the light intervals of the intermittent exposures are about equal to the average rate of incidence of the photons upon the silver halide crystals, the intermittency effect disappears; that is, the intermittent exposure yields the same density upon development as would a non-intermittent exposure of the same duration and average intensity." If I read that right, then a 20-second exposure at f/8 would give the same density as 5x 2 second exposures, evenly spread aross 20 seconds, at f/5.6.
It's also intriguing that with bright enough intermittent exposures, density increases as a result of the intermittency effect. This all seems to be related to reciprocity failure, which has two different roots: at low intensities, latent image regression or dissipation is a significant factor when compared with the speed of latent image formation, while at high intensities, there are so many free electrons that the latent image speck becomes negatively charged, repelling some electrons which disappear into the crystal and become unavailable for development; the latent image centres are not built up to a sufficient size.
At this point, we see that the problem here is intermittency effect almost certainly augmented by bulb warming. I'm not surprised when you say you don't understand the science: I've just been reading Haist's summary, and I strongly suspect that even if I read the original papers he cites, I still wouldn't understand it fully either. But I learned quite a bit while trying to find the answer, which is what it's all about.
Edit: I realize the above could sound arrogant: "Even I couldn't understand it, and I'm much cleverer than you are." What I meant was, "I'm very nearly as much in the dark as you are, even after reading it up in Haist."
I fully take your point that intermittency happens, and we have to live with it, but this doesn't remove the interest of finding out why and how it happens. In purely practical terms, it explains why we make work prints and then modify them, which is why I've always throught the 'maximum black' criterion to be of limited usefulness.
Finally, thanks for the reference to Strand.
Cheers,
Roger