Limitations of that sort are self-imposed:
There are a number of reasons why one might limit the length, and you've mentioned most of them.
Some others:
> IR drop
> bonder limits -- You have move the bonder tip up and out to draw additional length, and then come back down. I can't really if there really a mechanical limit on bonders, though.
> damage to first bond -- drawing additional wire puts stress on the first bond
I vaguely recall seeing bonds longer than 0.1" but that was mainly due to putting a really small die in a package that was too large for it. And that was due to the fact that the large package was off-the-shelf, and the exercise was for a special test.
As to some of the concerns you raised:
> droop -- no necessarily a problem, since die are usually passivated, so shorting to the die is unlikely. Shorting to other bonds might occur if the loop height is excessive. Certainly, the longer the bond, the more sway space you should have allowed for in the bond pad layout.
> fatigue -- unclear, since the longer the bond, the smaller the angular deflection relative to the foot of the bond.
My guess would be that the most prevalent reason is inductance, particularly with high-speed, high-transient current devices. A poor bonding layout can result in several tenths of volts of bounce during a transient, enough to potentially change the logic state of the signal on the wire.
TTFN
FAQ731-376