Architecturally the system you describe is very similar to the modern banking system with high powered 'peers' upon whom the users are utterly dependent. All that remains to be seen are what the ratio will be. It is simply disingenuous to refer to it as a 'peer2peer' solution and retain any real meaning to the term.
If you've ever tried to set up a bank you'd know there is no comparison possible. A bank is not just a computer that tracks transactions, you know. And you can't just switch from one to another in a second or two.
In the modern western world 'banks' (usually understood by the masses to be consumer banks) are corporations and like all corporations it is their legal obligation to maximize shareholder value. They happen to deal with currency as an artifact of their doing business, and in doing so make use of centralized computers which interact with the computers of their 'peers' and those of central banks and regulatory bodies. I'm just making this up, so please feel free to correct me if you feel that I've mis-spoken.
Other corporations maximize shareholder value in other ways, but visibility into the interactions inherent in a currency system can be extremely valuable to them. Hence significant interest in 'virtual currencies'.
One thing that does not at all maximize shareholder value is crossing the government or regulatory bodies, so it would be unwise to expect corporations to do so.
SPV mode is appropriate for people with hardware constraints that let them get almost-as-good security at much less cost. If you want the full security you can of course upgrade to use a full node.
As long as it is realistic for a fair fraction of the user base to be full peers I'm not concerned or unhappy. This is, to me, incompatible with unlimited or heavy growth. I am not super concerned about mining because to me seems to be something which can be isolated and the system will still remain 'open'. If latency increase dramatically (due to near-monopolization of mining power by corporations) that still will not preclude the use of Bitcoin as a reserve store of value.
My concern is that the solution will be grown enough such that full peers will require such specialized and expensive hardware and network connectivity that it will be practical for them to be isolated, controlled, and made inaccessible from large swaths of users through DPI and packet filtering. Eventually.
I am also concerned that the business intelligence value of being able to closely monitor economic activity in a dominant currency solution will far exceed the costs of operation yielding a situation where it is provided as a 'free' service in the same vein as we see with other forms of communication over the internet these days. This will greatly disrupt some of the economic incentives which are supposed to keep the system in balance. Or at least 'supposed' by me to be part of the design.
But that blockchain is a pain to download, and having to catch up every day takes a while. Some find it appropriate to sacrifice that "trust no-one" stasis for convenience. 
