I reported the reduced MWE (the one in post #26) to Stata tech support. Tech support replied:
To recap, Enrique's response was that this instability is desirable behavior because it indicates an identification problem. João says there is no identification problem. I don't see an identification problem, but maybe I don't know what I'm looking for.
I… I'm stumped. What should I do? Should I just assume that Stata's developers are incorrect, and live with the very minor inconvenience that this issue causes? I'm uncomfortable with that assumption.
I reported the issue to the developers and, after reviewing your new example, the conclusion was still the same provided by Enrique in his response to your Statalist posting.
I… I'm stumped. What should I do? Should I just assume that Stata's developers are incorrect, and live with the very minor inconvenience that this issue causes? I'm uncomfortable with that assumption.
Comment