ASIC resistant algorithms are not doomed to failure. The problem is they are implemented so poorly because they just have to cater to GPU's and therefore they make the memory requirement too small. Even monero made it small enough to fit in the processor cache. It is just not big enough. You may say well the asics will just add more memory, the problem is that random memory accesses take lots of processor speed, which GPU and ASIC need to minimize to be competitive. Scrypt, the classic alt algorithm only picked 1024 for its memory size (N value) where for real asic resistance you need 20,000 even up to 50-100,000 is no problem for CPU"s. You can't have your cake and eat it too, if you want GPU's to be fast at mining then you will have an ASIC problem, if you design it so GPU's will struggle and GPU miners won't like mining your coin, then you are safe. We need to mature as a community I feel. It is time to drop our GPU love affair. The ideal algorithm will require a CPU and GPU in tandem, and this algorithm is called "Factorization of large numbers" wherin an ASIC has never been created though a incentive has existed for decades. A miner for this algorirhm requires a CPU and GPU.
Right, there are actually quite a number of ASIC resistant one-way functions out there that really can go on to dis-incentiving Sybil attacks. The easiest solution is to look into mechanisms for anti-denial of service which I believe alluded to Bitcoin's choice for Hashcash originally purported for anti-mail spam.
Apart from Scrypt, there is the memory-hard function Argon2, and a whole paper titled "Asymmetrically and Symmetrically-hard Cryptography" from ASIACRYPT 2017 that denotes the enabling of resource-hardness through plugs in a wide variety of cryptographic hash functions.
The key to creating anti-DDoS mechanisms (such as the ones you'd see in node identity derivation is S/Kademlia) really is to make a function that is only really computable on general-purpose computing devices. Ones that are selective in their forms of resource-hardness only enable ASIC resistance for ever-so-long (as we could see out of Equihash) given an ASIC's capability in specialization towards computing very selective, specific tasks.
