If you want things in the block chain to represent shares (ie: you want the management to count your vote come agm time or pay you a dividend) or you want some farmer to deliver you some corn then they (the management or the farmer) becomes the point of failure. You're relying on them.. the fact that there's a block chain that says they should do something is completely irrelevant. They might as well just keep their shares/corn orders in a very ordinary spreadsheet.
That's true for corn, yes, but not stock, which can be traded electronically, automatically and securely.
And for corn, that is analagous to selling a coffee or a car for bitcoins. You are relying on an external point of failure.
So what if the corn (or coffee) is never delivered. Having one element outside the scope of the project does not eliminate the utility of the bitcoin approach.
No, you're still missing the point.
Stock is only useful if honored by the company management. If you have to trust them (which you do) they might as well just maintain the share register (which is exactly how it currently works, although most companies outsource the maintenance of the database).
There's no point in using a complicated proof-of-work system to maintain a simple spreadsheet when at the end of the day you're effectively just going to print off a copy and hand it to someone and just trust that they do whatever it says. They might as well maintain it centrally as it's much simpler. There's no value add for this huge p2p network crunching thousands of Ghash/s when at the end of the day someone can just ignore it.
The only reason the bitcoin network works for bitcoins is there are no external trust dependencies. That won't be true in the case of any of the physical goods trading you described.
BitDNS could be done that way, but it could also be done a much simpler way without proofs of work.