Stuck vcpkg lock: sccache daemon holds it

When a cmake configure hangs printing "waiting to take filesystem lock on …/vcpkg/.vcpkg-root", the holder is almost certainly the sccache daemon. vcpkg install takes an exclusive flock() on .vcpkg-root and compiles ports through sccache; if no server is running, the first call auto-spawns the daemon, which inherits the non-CLOEXEC lock fd. flock follows the open-file-description, so the daemon keeps the lock alive long after vcpkg exits — and it never idles out while other worktrees keep it busy.

Why: Hit on 2026-06-04: builds in OreStudio.local3 spawned the daemon while vcpkg held the lock; hours later a configure in the same worktree deadlocked while the daemon served compiles for OreStudio.local4. fuser=/=lsof showed nothing; only a /proc/*/fd scan found the holder.

How to apply: To unblock: sccache --stop-server (it restarts on the next compile; in-flight compiles in other worktrees may error once — re-run their make). To prevent: pre-start the server outside any build, e.g. sccache --start-server at login, so no build ever spawns it while holding the vcpkg lock. To diagnose a held flock that lsof misses, scan /proc/[0-9]*/fd for links to the lock file.