Recovery behavior for unknown route could be improved by generating new route cookie

[Former user]

Hi,

When a provided route contains a reference to an invalid (e.g. dead) peer, the current behaviour is fallback to round-robin (until the user closes his browser, which destroys the cookie).

I would prefer a behaviour where the mechanism selects another peer, and then writes a NEW route cookie referring to the new peer.

A typical scenario is that we kill a buch of instances after an update, and add several new instances to the pool. In the current situation all users are suddenly "stuck" in round-robin mode.

[Former user]

My apologies. Additional testing and studying of the source code proved the behaviour is already as I described in the issue.

[Lazy DogTown]

thanx fo r re-checking