A single design failure could render all the account information in the blockchain useless.
This is one of the very good reasons why the world will not, and should not, convert from the legacy system into a crypto-centric one instantaneously. We are not going to jump from here to there all at once or overnight. We are going to do it cautiously, taking a million baby steps in the process. There are scenarios for catastrophic failure that you have thought of, some that I have thought of, and I am sure some that neither of us have thought of. What if some crazy obscure bug causes the chain to fork into a thousand directions all of a sudden? But the thing is, we can play what-if for any technology. What if every computer on the planet gets knocked out by a gigantic EMP or series of EMPs, man-made or otherwise? All of our financial information would be lost! OOMMMGGGG!!! So, what do we do? Just abandon technology, because it might fail us? No, we are going to have to have backup plans in cases of catastrophic technical failure of the blockchain, just like we do for any new technology. I suppose I have more faith in our ability to do so than you. (Or not. I do not want to be presumptuous.)
A government could break all bitcoin-dependent businesses at once by blocking access to the bitcoin network (by network hacking or by legal threats on users).
Yes, and the government could shut down the internet, and the government could turn off your cell phone and cut off your electricity. And the government could shoot you in the head. All of these being arguments that make crypto more appealing, not less.
And it is undeniable that, for the amount of traffic that it processes, the bitcoin network is terribly more wasteful of computing resources than the existing financial system, by orders of magnitude.
Are you talking about the way things stand now, or the way things are envisioned to be in 2, 5, 10 years? Right now a lot of money is spent per transaction [1]; but that is basically because the infrastructure is still being built and people are investing lots of money. If you were an airline and were spending a zillion dollars to build a fleet of new airplanes, and only one airplane was in service so far, would you calculate the cost per passenger as the number of passengers flying right now in the one airplane, divided by all the money being spent building the entire fleet? No, you would not, because it would be a disingenuous argument to make.
[1]
https://blockchain.info/charts/cost-per-transaction?timespan=all&showDataPoints=false&daysAverageString=1&show_header=true&scale=0&address=