I am looking forward to Hornet[0]. The authors haven't put a PoC out yet. It is described decently in a recent mailing list post[1] (and impl[2]) on a BitCoin decentralizing project. A quality project might take the speed of the Hornet protocol, the tor hidden services approach, remove the exit node concept, make everyone a relay, and incentivize bandwidth sharing.
Sadly, I don't think we will witness a working implementation of Hornet anytime soon. Hornet was designed to be implemented at the core Internet infrastructure in order to operate at the network level. It is practically -impossible- to achieve this level of migration (remember IPv6). Moreover, it requires the deployment of a PKI on an unprecedented scale, which may or may not be secure. However, it's true that some ideas might be extracted and used in other future projects.
Wrong. The exit-node might control the traffic but that does not know where you came from. You need to watch the connection into the torrnetwork as well.
When you connect to https its pretty hard and risky for the exit node. If the user uses https with pinning its even harder and more risky.
The Tor Network is not perfect, nothing is in this space. So you need to make a more detailed analysis then yours to evaluate it.
0 - http://arxiv.org/pdf/1507.05724v1.pdf 1 - http://lists.linuxfoundation.org/pipermail/lightning-dev/201... 2 - https://github.com/LightningNetwork/lightning-onion