First of all, let's acknowledge that ASICs gets continually eliminated by the competition naturally as the rewards decreases. The network doesn't suddenly gets substantially insecure after the various halvings.

If the cost of executing such an attack gets low enough, that will probably mean Bitcoin isn't doing so well. Attacking it is pretty meaningless, we'll just fork it into another algorithm and the attackers don't earn anything and life goes on. ASICs are highly specialized and aren't that easy to manufacture and get your hands on. Both the electrical constraints, space constraints, hardware constraints will almost definitely persist in the future. These are legitimate hindrance to a successful attack. Purchasing so much ASICs and expending resources just to attack something that is half dead (if the network gets less secure) isn't very smart.

Try scaling the current network to any possible attacks, including the resources required, difficulty of obtaining them and the opportunity costs incurred by a state sponsored adversary or a very rich person.