First, the hashrate's primary purpose is not mining transactions, it's securing the network. That includes transactions that have already been mined, and includes all of the bitcoins that everyone is sitting on. For me personally, Bitcoin's primary purpose is not moving money around but rather parking my money safely in an asset that is not subject to the whims of politicians. Giant price swings are inconvenient, but I am willing to accept the variance because I know my money cannot be seized, my transactions cannot be censored, and the money will not lose value because my government decides to print a ton of cash. It is important to me that the network have a high hashrate whether or not I am actively moving my money around, and in fact most of that $11,000 is charged in the form of inflation, not in the form of transaction fees. Inflation penalizes the people holding onto their bitcoins, where fees penalize the people moving money around.
Secondly, the cost of a transaction is a relatively minor concern when it comes to the blocksize. What is most important is that the network is secure, and cheap transactions come as a secondary concern. If you want cheap transactions, you can use a centralized system which does not need to worry about all the problems that accompany decentralized money. If you want decentralized transactions, you're going to need to pay the minimum price necessary to make sure that the network remains both secure and decentralized, and (unfortunately) 1MB (soon to be ~1.75MB, thanks to segwit) per 10 minutes is about all we can do safely.
Explaining why 1MB is approx the best throughput the network can get is a much longer conversation, but there are signficant safety concerns related to hardforking the network, related to slowing down block propagation (larger blocks propagate slower, and if blocks are taking more than ~6 seconds to propagate the network you get mining centralization pressures), related to the cost of running a full node, related to the fact that a fee market is going to be what secures the network in the future, and related to a handful of other smaller issues as well. It turns out that a number as simple as 1MB has a large number of broad effects on the system as a whole, and changing it even slightly can have pretty dramatic effects in unintuitive places.
Finally, in practice there is not a linear correlation between hashpower and block subsidy. Lots of miners have been preparing for the halving for months, and therefore have only expanded their operations to what they will be able to sustain after the halving. There are at least a few large mining companies out there that are actually spending less than half the price of a bitcoin in electricity and maintenance for each bitcoin they mine, meaning they will not lose any hashrate at all when the halving arrives. There was some concern about this at the Scaling Bitcoin conferences, so me and a few others probed deeper into the probable outcomes of the reward halving, and we found that most miners had a plan which did not involve them turning off their rigs even if the price did not rise. Some of them are actually already locked into power agreements that will penalize them heavily if they don't consume the electricity.
It's difficult to predict what the halving will do to the price. We know that when the halving occurs, ~1800btc per day of sell pressure will disappear. That most likely means that the price will rise, but there is absolutely no reason to believe that this will cause the price to double.
I personally am predicting that, regardless of price movement, we will see the hashrate drop less than 15% in the month following the halving. I have done a moderate amount of research in arriving at this conclusion, at least more than your typical armchair speculator.