0.15.1 — Parallel Daemons · Kandown changelog

Fixed: Multi-project daemon port clash — launching kandown from a second project folder no longer kills the first project's web daemon and steals its port. Each project now gets its own daemon on an auto-allocated port (A=2048, B=2049, C=2050, …), so multiple kandown boards can finally run in parallel.
- Root cause #1: listenOnAvailablePort() unconditionally killed whatever kandown sat on the default port — even when it belonged to a different project. The preemptive stale check now only reclaims same-project zombies; other-project daemons fall through to the port-scan loop (which already skips them).
- Root cause #2: the parent↔child startup handshake deleted the freshly-written daemon metadata on any transient HTTP fetch failure. Node's fetch (undici) doesn't recover from the initial ECONNREFUSED window and reported healthy local daemons as down for seconds, orphaning them with a spurious "Daemon failed to start". getDaemonStatus() is now non-destructive on transient fetch failures (metadata is removed only on a real conflict), and startup detection uses a dependency-free TCP probe (net.createConnection) instead of HTTP.
- Port range widened 2048–2060 → 2048–2150 (103 slots); startup timeout 5s → 8s.