Is it not a benefit than scrypt is less susceptible to those big established mining networks?
Or would scrypt invariably end up the same? I know for sure ASICs are not a possibility, right?
Only if you define benefit as the temporary absence of a weakness.
Of course there can be an ASIC for scrypt. For one, all you need to do is to minimize the number of transistors used for the core algorithm, getting rid of all the expensive instruction handling in universal processors... At least that will increase computational efficiency. It may take a long time to ROI though, dependend on the market cap (capital expenditure ROI is a function of the opportunity cost difference between acquiring the good vs. the productive capacity).