Never did it on my NF7-S, but I was still able to get 245 FSB rock-stable, which was probably the limit of the board's northbridge. I ran TicTac Mantaray XT D26 BIOS, which has a "soft L12" mod, and I ran the Abit D26 and D27 BIOSes, and it made very little difference. This was a mobile 2600+, by the way, so default FSB was 133. It seemed to have no problem running at almost twice that, and the "soft L12" mod didn't seem to make any difference either.
Personally, I think it's just one of those things where maybe people reseat the chip and heatsink to make better contact, and get higher FSB as a result of improved heat transfer/cooling. Or maybe people just want to believe that it makes a difference, so they don't test stability as thoroughly, and just assume they suddenly gained an extra 20 MHz of FSB. From a logical standpoint, it makes zero sense that some people claim shorting the bridges to make 133 FSB CPU's use 166 FSB get improved performance. There are also people now claiming that doing the opposite, setting 166 FSB CPU's to default to 133 FSB, has increased their FSB. This goes against all logic, and is probably more board specific than anything.
The only thing I can possibly think to explain this is that some boards might have some sort of "safety feature" that is switched on when FSB is set too high above default, and they refuse to boot. That would explain how setting FSB to 166 would give you up to 33 extra MHz, but it doesn't explain how changing 166 FSB to 133 default would change anything. Obviously, the CPU itself doesn't care about what its default FSB is, and it doesn't "prefer" one over the other. It is inherently capable of a maximum FSB depending on the quality of the chip, and there is no logical explanation of how jumping pins or bridges would suddenly turn the silicon in the core into higher-quality and capable of higher FSB.