We saw a similar thing with RUCoin (which had SHA256 and scrypt) and the coin failed.
If you're saying weighting can cause a big disturbance from a minor shift in hashing, that phenomenon seems much less likely to occur with 4 algos instead of 2 if weighted properly.
It's obviously also much more cartel proof.
People like Gmaxwell seem to claim that a group like KNCminer would create ASICs for all four chains.
I don't really see groups like that monopolizing all four chains into eternity, especially if each algo is radically different from one another.
They aren't radically different. All you need to implement is an instruction cache that does most of the normal transformations for hashing algorithms (eg bit rotations). Then you can make an ASIC for any hashing algorithm that can be implemented in 20k-100k circuits. scrypt was neat in that it required some memory, but then they made ASICs with memory.