> In the discussion that you cite at the beginning of this bug report,
> I suggested a different approach to finding out the culprit, by
> using the redisplay tracing facility. Can you try that and post the
> results? It might give better clues.

Though I can't reproduce it at will, I do have a trace-redisplay from
when I saw the bug in a Java file. Specifically, I saw point move
inappropriately, but redisplay didn't need to scroll. It's attached in
case you find it useful.