Reactor Scram
Reactor Scram
I split off the headless client since that won't be part of beta, so all remaining tasks are complete, we can treat any new tasks as new issues
list of routes does sound nice. Then the only diffing is in the client, or they could blow up the routing and re-do it, as a simpler implementation. Adding 1...
When I copy the non-blocking code from the headless client this might coincidentally get fixed. And I may make a debug subcommand to replicate it without having to sign in...
Still happening with non-blocking
Here it is replicating on 679bd827264b0, on WSL Linux. I modified it to run the Session for 20 seconds, then reconnect. Without sending any trafffic (good)  If I send...
Set up tokio-console again. It's those same 3 tasks I think. Captured on 99e700b80 
The webrtc repo says there might be some memory leak / Arc reference cycles, since the code was ported from Go, which has a GC that can break ref cycles...
Discussed in standup. We may put this on pause. - `firezone-connection` might fix this incidentally soon https://github.com/firezone/firezone/tree/main/rust/connlib/connection - We can still stick it in a subprocess if we need
#3259 might be related to this
Well I didn't merge the IPC branch into firezone/firezone, so if we want to ship without firezone-connection I'll have to do that