Some protocols have experimented with dummy traffic (see the survey). How will these affect that?
At random, drop an incoming onion.
At random, create a route to a random node where the last layer is corrupted (hence that node will drop it).
I don't think any of this will help. Even if you forward a dummy packet every time you receive something to yourself, an abnormal amount of packets will pass through you (more than a uniform distribution).
Since the destination is a set, we will not exhibit the same problems as a random walk. However, if only a subset is online, then we approach the random walk distribution.
Some protocols have experimented with dummy traffic (see the survey). How will these affect that?
I don't think any of this will help. Even if you forward a dummy packet every time you receive something to yourself, an abnormal amount of packets will pass through you (more than a uniform distribution).