Never read sysfs through Go's poller, and a GPU that stops answering is skipped

On grid the sampler sat 42 minutes in runtime_pollWait on the Arc A380's
hwmon fan1_input: kernfs accepts epoll, so os.ReadFile made the fd
non-blocking, i915 answered EAGAIN once, and the poller waited for a
readiness kernfs only signals on sysfs_notify. Every graph on the machine
froze while the API carried on.

readStr now reads with plain blocking syscalls. Each GPU's sysfs reads run
under a three-second deadline; a device that does not answer is marked
stalled, logged, and left alone until restart, with the box saying so.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

internal/metrics/metrics.go

@@ -1352,12 +1352,32 @@ func Signal(pid int, sig syscall.Signal) error {
 
 // ---- small helpers ----
 
+// readStr reads a /proc or /sys file whole, with plain blocking syscalls and never through
+// the runtime poller. os.ReadFile would hand a sysfs file to the poller — kernfs accepts
+// epoll — and a driver that answers a non-blocking read with EAGAIN (i915's fan attribute
+// did, on grid, 2026-10-02) then leaves the reader waiting for a readiness kernfs only
+// signals on sysfs_notify: never. The sampler sat in that wait for 42 minutes.
 func readStr(p string) string {
-	b, err := os.ReadFile(p)
+	fd, err := syscall.Open(p, syscall.O_RDONLY|syscall.O_CLOEXEC, 0)
 	if err != nil {
 		return ""
 	}
-	return string(b)
+	defer syscall.Close(fd)
+	var out []byte
+	buf := make([]byte, 4096)
+	for {
+		n, err := syscall.Read(fd, buf)
+		if n > 0 {
+			out = append(out, buf[:n]...)
+		}
+		if err == syscall.EINTR {
+			continue
+		}
+		if err != nil || n == 0 {
+			break
+		}
+	}
+	return string(out)
 }
 
 func readInt(p string) (int64, bool) {