In this case you should use I2P, which is specifically designed for hidden services (eepsites) first.
I2P has a completely different architecture and AFAIK it remains to be known which one is better.
Isn't it obvious that when everyone is a middle node by default and all tunnels are one-way and randomly changing every 10 minutes, it must be harder to perform deanonymizing attacks?
On the other hand, one tunnel is one user for 10 monutes at a time. And everyone knows when your client is online because it's a relay.
Yes, just like with Tor, the fact that you're using it is not a secret. What you're doing there is.
If you're a relay, then the dates and times you are online are public and may be correlated to the times a particular username logs into a particular hidden service, if that is also public. With Tor, I'd at least have to snoop your specific connection or your guard relay to know when you're online, but with I2P, it's public knowledge.
True. Although you can opt out of being a rely on I2P, or by adding more hops you will make such correlations less reliable (together with the reliability of your connection).
Also AFAIK only your destination eepsite can perform such an attack, which is not very scalable for an attacker.