Earlier quoted context omitted.
Don't bitcoin costs adjust to a shrinking miner population as well as it adjusts for a growing one? If so, the answer is never. As people give-up, the costs for the remaining ones reduce so that at some point it's lucrative again. The interesting question is at what price people will stop speculating on its price and run to an exit? I don't think anybody can answer this one.
The difference is that even if the difficulty adjusts, there will be dormant ASICs waiting to be used. If the price lowers enough and enough ASICs are out on the bench, it’ll at one point becoming economically beneficial to turn them all on and attack the network rather than having them collect dust and depreciate.
Getting those ASICs from where they are unused to where someone can use them to attack the network is not a simple task. The only way this becomes possible is if the current owner of the hardware decides to attack the network. Even to buy the hardware will require a large capital input, let's at you can buy $1m worth of hardware and ship it to your location within 7 days (shipping and setup), you then need to run it (electricity cost) and hope that you can recoup costs plus profit by double spending. The only option to double spend and profit is to double spend on an exchange and withdraw fiat currency before the exchange realises, the speed at which fiat transfers work this is highly unlikely.