fix(stream): don't cache transient libplacebo probe timeouts

Second critico pass on the functional probe.

- The probe does real Vulkan device init, which can transiently fail when the
  box is busy (notably the startup warm racing the encode benchmark). Caching
  that timeout as a permanent 'no' would pin HDR to the zscale CPU chain until
  daemon restart. Now a deadline is NOT cached — only a clean non-zero exit
  (filter absent / no ICD), which is a stable result. zscale stays cached as
  before (cheap deterministic grep, can't flake).
- Surface the exec error when ffmpeg never produced stderr (timeout / ENOENT):
  the fallback log now shows err.Error() instead of a blank tail, so 'no
  Vulkan' is distinguishable from 'ffmpeg never ran'.
- Dockerfile comment: clarify the Vulkan ICD (not GLX) is the load-bearing
  mount that 'graphics' adds; 'compute' alone doesn't mount it.

Probe still returns true on a Vulkan host (verified); engine tests green.

Dockerfile

@@ -95,9 +95,10 @@ ENV XDG_DATA_HOME=/data
 
 # NVIDIA passthrough defaults. `--gpus all` alone only grants the "utility" +
 # "compute" capabilities; nvenc needs "video", and "graphics" makes the runtime
-# mount the NVIDIA Vulkan ICD (nvidia_icd.json + GLX libs) so ffmpeg's libplacebo
-# filter (GPU HDR tonemap, paired with libvulkan1 above) can create a Vulkan
-# device. Baking these here means a plain `docker run --gpus all` (or the compose
+# mount the NVIDIA Vulkan ICD (nvidia_icd.json — the load-bearing piece — plus
+# GLX/EGL libs) so ffmpeg's libplacebo filter (GPU HDR tonemap, paired with
+# libvulkan1 above) can create a Vulkan device. "compute" alone does NOT mount
+# the ICD. Baking these here means a plain `docker run --gpus all` (or the compose
 # device reservation) lights up HW transcode + GPU tonemap with zero extra flags.
 # Harmless when no GPU is attached.
 ENV NVIDIA_VISIBLE_DEVICES=all

internal/engine/tonemap.go

@@ -57,11 +57,12 @@ var (
 // tonemapped fine via zscale.
 //
 // So we run the real filter on one synthetic frame and require a clean exit:
-// that forces Vulkan device creation + filtergraph negotiation (the implicit
-// hwupload/hwdownload around the GPU filter). Pass → libplacebo works here;
-// fail → fall back to the zscale chain. Cached per path; a probe failure is
-// treated as "no". The probe is bounded so a wedged ffmpeg can't stall the
-// first session.
+// that forces Vulkan device creation + filtergraph negotiation (libplacebo
+// auto-inserts the hwupload/hwdownload around itself). Pass → libplacebo works
+// here; fail → fall back to the zscale chain. Cached per path — EXCEPT a
+// context timeout, which is transient (a busy box during the startup warm) and
+// must not pin HDR to zscale for the whole process. The probe is bounded so a
+// wedged ffmpeg can't stall the first session.
 func FFmpegSupportsLibplacebo(ffmpegPath string) bool {
 	if ffmpegPath == "" {
 		return false
@@ -87,13 +88,27 @@ func FFmpegSupportsLibplacebo(ffmpegPath string) bool {
 	).CombinedOutput()
 	supported := err == nil
 
-	libplaceboCacheMu.Lock()
-	libplaceboCache[ffmpegPath] = supported
-	libplaceboCacheMu.Unlock()
+	// Cache the result — but NOT a timeout. A clean non-zero exit (filter
+	// absent, no Vulkan ICD) is a stable "no" worth remembering; a deadline is
+	// transient (the box was busy, e.g. the startup warm racing the encode
+	// benchmark) and caching it would force HDR onto the zscale CPU chain until
+	// restart. Worst case a perpetually-loaded box re-probes per session — rare,
+	// and it fails closed to zscale each time.
+	if supported || ctx.Err() != context.DeadlineExceeded {
+		libplaceboCacheMu.Lock()
+		libplaceboCache[ffmpegPath] = supported
+		libplaceboCacheMu.Unlock()
+	}
 	if supported {
 		log.Printf("[tonemap] ffmpeg libplacebo works (Vulkan OK) — HDR sources tonemapped on the GPU (preferred)")
 	} else {
-		log.Printf("[tonemap] ffmpeg libplacebo unavailable (no Vulkan runtime or filter absent) — HDR falls back to zscale/none: %v", strings.TrimSpace(lastLine(out)))
+		// On an exec/timeout failure the stderr tail is empty — surface err
+		// itself so the log distinguishes "no Vulkan" from "ffmpeg never ran".
+		detail := strings.TrimSpace(lastLine(out))
+		if detail == "" {
+			detail = err.Error()
+		}
+		log.Printf("[tonemap] ffmpeg libplacebo unavailable (no Vulkan runtime or filter absent) — HDR falls back to zscale/none: %v", detail)
 	}
 	return supported
 }