> That's a workaround

No, from dictionary "A workaround is a bypass of a recognized problem in a system", I didn't know about the MultiPV bug, so didn't know it had a problem. I'd rely on exclude moves with or without it (basically it doesn't affect me).

> but hard to say it's better than simply using multi-pv.

I'd turn it around and say that it's hard to say if using MultiPV is better than simply excluding moves.

> I don't know how you would use exclude moves for IA without having an IDeA tree first

Where do you store MultiPV analysis? The hash is the worst possible place as that will get deleted on next load or overwritten soon enough. So people have to learn to store their analysis later for when it's eventually needed, like, 3 months later when a different opponent reaches the same position that was already analyzed, running MultiPV again is pointless if you already did that back then, it's much more effective if you just look at what you already have stored, AND you can use exclude moves on all analyzed moves to look for new ideas. I don't use IDEA trees, but I've heard about people that store their analysis even in text files.

Bottom line, MultiPV seems like outdated technology that doesn't have any use, specially with Rybka, that is slowed down more than she should, try with Naum, tick the "Preserve analysis" box, put it to analyze the position to some depth, exclude the main move, and in transpositional positions, it'll immediately show another move with the same depth, that you can exclude to get a third move at that depth, all instantly, in MultiPV=3 it'd take about 50% more time to get the same output, so I think this R4 bug is an excellent chance for people to learn how to use superior analysis methods like exclude moves.

This happens frequently to me when I do backwards analysis, but it can't be reproduced easily because it is dependent on what you have in the hash. Clearing the hash solves the problem at the cost of throwing out all the time you spent on putting in best known lines.

Yes, all Persistent Hash bugs had its workarounds, just for that Rybka 3 is still a core engine of mine. Rybka 4 is my fourth main engine, I'd be crazy if I put it as the first one. I have this feeling that some people are probably just doing weaker analysis and performing worse by relying on R4 instead of R3, but they're not aware of it and assume that just because it's 40 elo stronger on the rating lists, that it'll automatically give a better job.