I suspect what happened here - and which is currently happening with Ethereum, IMO - is that the Monero devs just couldn't believe an ASIC miner was practical - it is always possible, yes, but not necessarily worth doing. So when network hashrate started going up dramatically a few months ago they likely attributed it to their burgeoning success rather than an ASIC. It was only when Baikal, etc., showed videos of their ASIC CN miners that the threat even registered, but by then they were months behind the curve.

Same thing with Ethereum - the devs claim to hate ASICs six ways to Sunday and that the heavy use of memory by the Ethash algo makes an ASIC miner impractical, yet the Bitmain E3 just went on pre-sale for July delivery and promptly sold out. Seems that Bitmain found out that paralleling a few channels of cheap DDR3 was just as good as a single channel of DDR5 and the rest is about to become history.

These two examples should serve as a prominent warning to all other coin devs: if you want to resist ASICs taking over your network then you need to preemptively fork every 6 months, at least; if you merely want to prevent decentralization of your network because of a single ASIC manufacturer, then figure out the HDL implementation of your algo and release it as public domain so many new FPGA and ASIC miners can be developed rapidly.