I don't think any "ASIC resistant" algorithm would be desirable to implement. All it would do would be to make it more difficult to develop ASICs, and there is always the risk that the person advocating for a particular algorithm has already done some research on how to create ASICs for that algorithm.
So far based on what I've seen ASIC resistant meant you can't use bitcoin SHA256d ASICs to mine that algorithm but creation of an application-specific integrated circuit (ie. ASIC) is very well possible. Even those that tried using more memory with memory expensive algorithms ended up having ASICs.
The only working solution so far seems to have been changing the algorithm every now and then!
Right. ASIC resistant algorithms have only delayed the creation of ASICs.
P.S. Do we even want an ASIC resistant algorithm? After all ASICs with their huge hashrate are providing the security.
I don't think having a high hashrate is necessarily something that provides security. If the hashrate is 1000 hashes per second, an attacker will need 500, or 50% of the hashrate to reliably attack the network. If the network hashrate is 100k hashes per second, an attacker would still need 50.01% of the network hashrate to attack.
What I believe provides more security is the value of the miners compared to the value of the miners if there was no bitcoin, and the R&D cost of developing the miners. If there is a high cost to develop the miners, and if the miners are valuable, the cost of attacking the network will be very high.
The argument against ASICs is that ASICs will lead to there being fewer people mining because of the high cost of ASICs. So there is the risk that a small number of miners could potentially team up to attack the network. Without ASICs, the mining process would be something closer to "one person, one vote". I think if there were no ASIC (or GPU) mining, it would still be possible for an attacker to spin up 100k VPS on GCS or AWS to attack the network.