agent (unreleased): fixes after an independent review — one lock for the writers of agent.json, the trust run serialised and bounded, the CA copy keeps its previous root, the profile stanza from the wrapper's answer, a stale give-up ends, both boot tokens from one read

Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS

internal/pvetrust/pvetrust.go

@@ -25,6 +25,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/config"
 	"log/slog"
 	"os"
 	"path/filepath"
@@ -81,12 +82,16 @@ type Manager struct {
 	// MinGap between two runs that a failed call asked for (default 10 minutes).
 	MinGap time.Duration
 
+	runMu   sync.Mutex // serialises Ensure
 	mu      sync.Mutex
 	cur     proxmox.TLSConfig
 	status  Status
 	lastRun time.Time
 }
 
+// copyStepTimeout bounds the root step that makes the CA copy.
+const copyStepTimeout = 90 * time.Second
+
 func (m *Manager) now() time.Time {
 	if m.Now != nil {
 		return m.Now()
@@ -207,6 +212,11 @@ func causeText(cause string) string {
 // Ensure runs the order of the package comment once. why is a log word ("start", "daily", a trust cause). It returns
 // true when the running trust changed.
 func (m *Manager) Ensure(ctx context.Context, why string) bool {
+	// One run at a time. Ensure is called from the start, the timer, a failed collect and after a bundle install; two
+	// overlapping runs would both read the config, and the second would park the ALREADY-MOVED file as the "before"
+	// copy (losing the fingerprint from the recovery copy). Found by review, 2026-11-14.
+	m.runMu.Lock()
+	defer m.runMu.Unlock()
 	m.mu.Lock()
 	cur := m.cur
 	m.lastRun = m.now()
@@ -230,8 +240,11 @@ func (m *Manager) Ensure(ctx context.Context, why string) bool {
 	if m.EnsureCopy == nil || m.Probe == nil {
 		return false
 	}
-	// 1. the copy
-	c, err := m.EnsureCopy(ctx)
+	// 1. the copy. The root step gets a deadline of its own: Ensure runs on the daemon's start path and inside the
+	// report loop, and a wrapper that hangs (sudo, a stuck read of /etc/pve) must not stop either.
+	cctx, ccancel := context.WithTimeout(ctx, copyStepTimeout)
+	c, err := m.EnsureCopy(cctx)
+	ccancel()
 	if err != nil {
 		lg.Warn("pvetrust: the root step gave no CA copy — the trust stays as it is", "err", err)
 		set(modeOf(cur), ProblemCopyUnavailable, truncate(err.Error(), 240), nil)
@@ -341,6 +354,8 @@ func truncate(s string, n int) string {
 // writeConfig sets proxmox.tls.ca_file and clears proxmox.tls.fingerprint in ONE write of agent.json. Every other key is
 // kept byte for byte. A recovery copy is parked in the state folder first; the result is read back and checked.
 func (m *Manager) writeConfig(caPath string) error {
+	config.FileMu.Lock() // one writer of agent.json at a time in this process (config.FileMu)
+	defer config.FileMu.Unlock()
 	raw, err := os.ReadFile(m.ConfigPath)
 	if err != nil {
 		return err