On cranky-toaster-86 (macbookair2013, omarchy, hydraheadflatscreen v2.2.4) body discovery kept failing with 'no reachable bodies' for chunky-turnip-23: both the LAN and WireGuard probes timed out at exactly the 1s direct-dial limit. Meanwhile every fresh process on the same machine (root and cederik, curl with --connect-timeout 1) reached 10.10.100.16:47990 instantly. The agent had been running since Aug 18 15:46, likely predating the hydraguard-air interface/route setup. Restarting hydraheadflatscreen.service fixed it immediately and the stream started end to end.
Suspected: a long-lived process that outlives network topology changes (WireGuard interface creation) keeps failing dials that fresh processes complete. Also worth raising the probe direct-dial timeout above 1s for cold WireGuard paths.
Repro/diagnosis trail and fix: session notes 2026-08-19