I know I can move the hold bones or make more constraints, but I'm suggesting a feature that would make that unnecessary.
The octopus example would be extremely cumbersome to make with methods you suggested. I know it's a silly example, but I think the issue of wanting to change which bones are constrained by a transform constraint is not uncommon.
Perhaps this will be a better example:
I'm making a dodgeball animation, in which my character has different kinds of balls flying towards them and they need to catch each ball and throw it back. There's a basketball, a baseball, a softball, a tennisball, a golfball, and a bowling ball (6 different balls).
With the way Spine works now, I would have to have one transform constraint for each ball. If I had a second character, I would have to have another constraint for each ball. If I wanted each character to be able to catch each ball with either their left or right hand, I'd have to have another constraint for each player for each ball. This would mean 6balls
* 2characters
* 2hands
= 24constraints
just to make each ball be able to attach to any hand.
If I were able to key which ball(s) was(were) constrained to each hand, I would only need one transform constrain per hand, 4constraints
. I would then just key the constrained bone to the bone that corresponded to the ball the character needed to catch at that time. It would also be easy to make the character catch more than one ball at a time by constraining more than one ball at a time. It would even be easy to make the character catch a ball with both hands by making one hand constrained to the other hand.
A similar example is characters that are trying on different hats.
Note that all balls (and all hats) must be visible at one time, so skinning different balls (and hats) is not a solution.