Originally Posted by
vredstein
If I understand it correctly, the difference would be most noticeable when you're running a derailleur with sloppy pivots, weak return spring, or worn chain. If you take a length of a new quality chain and flex it to one side, you'll get a certain amount of movement. If you take an equal length of a worn or poorly spec'd chain and flex it to one side, you'll get more movement.
On a smaller scale, this movement is occurring between the upper pulley and cog in the length of chain that is not contacting the pulley or cog, maybe two or three pin's worth. If a b-adjustment can decrease this free chain length by a pin or two, that slop is decreased or eliminated. The derailleur movement is now working to move a stiffer length of chain instead of wasting movement to take up the small slop. In many cases, this can mean the difference between a chain hesitating or even refusing to make the shift.
Not true, it affects performance regardless of the state of wear of drivetrain components.
Please read page 32-9 chapter 32 of barnetts for a more and thorough explanation of how b-tension and chain length affect shift performance. And why it is *essential* to shift performance that both be set as optimal as possible (for the latter it's the shortest chain length).
32-25 of the "demo" chapter here:
http://www.bbinstitute.com/dl/dx_demo_chapter_32.pdf
The demo chapter only has a small subset of that section that appears in the full version - but that gives you a small taste.