If more effective GPU miner will be released, simple changes in algorithm will make it completely unusable. For example, we can use bcrypt() result as additional salt. I.e. add stage in hashing process.
Current algorithm is:
1) Calculate B = SHA256(A)
2) Calculate C = scrypt(B)
3) Calculate D = SHA256(B, C)
4) Return D
Probably future algorithm:
1) Calculate B = SHA256(A)
2) Calculate C = scrypt(B)
4) Calculate D = bcrypt(C)
3) Calculate E = SHA256(B, D)
4) Return E
Like scrypt(), bcrypt() is also using many pseudo-random operations with memory during encryption process.
1) Well i'm sure there is NO WAY to create an Algorythm that will NEVER better implemented on some piece of future hardware.
2) LTC is used by bot-nets to generate some cash, and aslong it is in this situation it will never be accepted as "stable" currency.
3) The limits on how good current Algoritm can be implemented on GPU i've calculated somewhere above, on common hardware best can be Nvidia GT 590 with smthg like 1.2 MH/s (unreachable maximum). Is this that dangerous for LTC ? i think NO.
I think we should leave it atm as it is and spare afford for "fixing" situation for other more dangerous situation.
P.S. i think GPU mining for LTC can even be good for LTC stabilizing network hashrate. My idea is do nothing.