We had experimented with self generated id codes (SGICs) years ago but was kind of 'noisy'...

You can create IDs that way but that might be an issue if the surveys are supposed to be anonymous (or at least not so easily identifiable).

That's interesting. I had written something before on the comparability of logit model results but did not think about it in a measurement context...

Haha! Your books will thank you (it was easy to adjust it though, just add another book)!