3) To make Bitcoin less secure you'd need an alt that shares its PoW algorithm and is about as profitable as BTC.
Reducing the hashing power pointed at BTC only makes sense if there's a viable alternative to point it towards to. Otherwise you just have expensive paperweights and lose out on millions of dollars each day. This makes withholding hashing power just for the sake of trying to pull off a 51% attack somewhere later down the road rather unprofitable. And those ASICs are the only machines of potentially doing such an attack. And those ASICs that you would be withholding, will already be obsolete in 10 months' time the latest, so time is playing against you.
Put differently: In practice less hashing power only means less security if there's other hashing power somewhere hidden away that could try to pull of an attack. Bitcoin's hashrate 2014 was much smaller than it is today, yet in practice the network was just as secure.
Miners getting shut down because they are not profitable anymore is a regular part of the process. It's been happening for years every time a new generation of miners hit the market.
[...]
I haven't followed BCH's scaling plans, but if they really want to handle 50,000 transactions per seconds using on-chain scaling that would require the equivalent of 7,142 MB blocks per 10 minutes -- or say, 714 MB blocks every minute. Now even assuming they'd manage to compress transactions down to, say 20% in size -- something that BTC could likely also do, if such a solution were to exist -- we'd still be looking at 1,428 MB blocks every 10 minutes. Unless they fundamentally change the way Bitcoin works this is not going to end well.
Of course BCH's scaling approach could turn out to be more viable. And of course that could reflect in price and thus hashrate. Right now, however, BCH is -- by your very own definition -- incredibly insecure. It oscillates at only around 10% of global SHA256 hashrate, making it an easy target for double spend attacks should BTC miners collude and decide to do so. And unless BCH's scaling approach blows BTC's scaling approach out of the water -- which I personally highly doubt -- that won't change anytime soon.