A path with an apostrophe broke the remote shell quoting and a big file outran the connect timeout, so intact backups were reported corrupt and absent

Monitors/backup_verify.sh

@@ -122,6 +122,7 @@ source "$SCRIPT_DIR/../load_config.sh"
 parse_args "$@"
 
 SSH_TIMEOUT=15
+BACKUP_VERIFY_MD5_TIMEOUT_MAX="${BACKUP_VERIFY_MD5_TIMEOUT_MAX:-600}"
 
 # ==============================================================================================
 # ━━━ Setup ━━━
@@ -228,6 +229,7 @@ TOTAL_CHECKED=0
 TOTAL_MATCH=0
 TOTAL_MISMATCH=0
 TOTAL_MISSING=0
+TOTAL_UNVERIFIED=0
 SHARES_WITH_ISSUES=()
 
 for share in "${VERIFY_SHARES[@]}"; do
@@ -265,6 +267,7 @@ for share in "${VERIFY_SHARES[@]}"; do
     SHARE_MATCH=0
     SHARE_MISMATCH=0
     SHARE_MISSING=0
+    SHARE_UNVERIFIED=0
 
     for local_file in "${SAMPLE_FILES[@]}"; do
         [[ -z "$local_file" ]] && continue
@@ -276,19 +279,49 @@ for share in "${VERIFY_SHARES[@]}"; do
             continue
         fi
 
-        # Remote checksum via SSH — timeout protected
-        remote_md5=$(timeout "$SSH_TIMEOUT" ssh -i "$SSH_KEY" \
+        # The path is interpolated into a remote shell command, so it must be escaped for
+        # reuse as one word. A bare '$local_file' inside single quotes breaks on the first
+        # apostrophe — "Frieren - Beyond Journey's End" ended the quote early, md5sum fell
+        # back to reading stdin, and the empty-input hash d41d8cd9... was reported as a
+        # MISMATCH against a file that is byte-identical on the remote.
+        printf -v remote_q '%q' "$local_file"
+
+        # Existence and content are separate questions. Asking them together means a slow
+        # checksum is indistinguishable from an absent file.
+        remote_exists=$(timeout "$SSH_TIMEOUT" ssh -i "$SSH_KEY" \
             -o ConnectTimeout="$SSH_TIMEOUT" \
             -o StrictHostKeyChecking=no \
             root@"$REMOTE_SERVER" \
-            "md5sum '$local_file' 2>/dev/null | awk '{print \$1}'" 2>/dev/null)
+            "test -f $remote_q && echo yes" 2>/dev/null </dev/null)
 
         (( TOTAL_CHECKED++ ))
 
-        if [[ -z "$remote_md5" ]]; then
+        if [[ "$remote_exists" != "yes" ]]; then
             warn "$ICON_ERROR MISSING: $(basename "$local_file")"
             (( SHARE_MISSING++ ))
             (( TOTAL_MISSING++ ))
+            continue
+        fi
+
+        # md5sum of a multi-GB file cannot finish inside a connect-sized timeout. Budget by
+        # size — a 5.9GB file needs ~30s and was being killed at 15s, then counted MISSING
+        # even though it was present and correct.
+        local_size=$(stat -c%s "$local_file" 2>/dev/null || echo 0)
+        md5_timeout=$(( local_size / 52428800 + SSH_TIMEOUT ))
+        (( md5_timeout > BACKUP_VERIFY_MD5_TIMEOUT_MAX )) && md5_timeout=$BACKUP_VERIFY_MD5_TIMEOUT_MAX
+
+        remote_md5=$(timeout "$md5_timeout" ssh -i "$SSH_KEY" \
+            -o ConnectTimeout="$SSH_TIMEOUT" \
+            -o StrictHostKeyChecking=no \
+            root@"$REMOTE_SERVER" \
+            "md5sum $remote_q 2>/dev/null | awk '{print \$1}'" 2>/dev/null </dev/null)
+
+        if [[ -z "$remote_md5" ]]; then
+            # Present but unreadable within budget. Reporting this as a mismatch or a miss
+            # would be a claim the run did not earn.
+            warn "$ICON_WARN UNVERIFIED (checksum timed out after ${md5_timeout}s): $(basename "$local_file")"
+            (( SHARE_UNVERIFIED++ ))
+            (( TOTAL_UNVERIFIED++ ))
         elif [[ "$local_md5" == "$remote_md5" ]]; then
             log "MATCH: $(basename "$local_file")"
             (( SHARE_MATCH++ ))
@@ -303,8 +336,8 @@ for share in "${VERIFY_SHARES[@]}"; do
     done
 
     # Per-share result — only visible if issues found
-    if [[ "$SHARE_MISMATCH" -gt 0 || "$SHARE_MISSING" -gt 0 ]]; then
-        warn "$SHARE_NAME — match: $SHARE_MATCH  missing: $SHARE_MISSING  mismatch: $SHARE_MISMATCH"
+    if [[ "$SHARE_MISMATCH" -gt 0 || "$SHARE_MISSING" -gt 0 || "$SHARE_UNVERIFIED" -gt 0 ]]; then
+        warn "$SHARE_NAME — match: $SHARE_MATCH  missing: $SHARE_MISSING  mismatch: $SHARE_MISMATCH  unverified: $SHARE_UNVERIFIED"
         SHARES_WITH_ISSUES+=("$SHARE_NAME")
     else
         echo "$SHARE_NAME — all $SHARE_MATCH files match ✅"
@@ -325,10 +358,11 @@ echo "$ICON_VERIFY  Checked:   $TOTAL_CHECKED files"
 echo "$ICON_TIME    Duration:  $(format_duration $(( END - START )))"
 echo ""
 
-if [[ "$TOTAL_MISMATCH" -gt 0 || "$TOTAL_MISSING" -gt 0 ]]; then
-    echo "$ICON_SUCCESS Match:     $TOTAL_MATCH"
-    warn "Missing:   $TOTAL_MISSING"
-    [[ "$TOTAL_MISMATCH" -gt 0 ]] && echo "$ICON_ERROR   Mismatch:  $TOTAL_MISMATCH"
+if [[ "$TOTAL_MISMATCH" -gt 0 || "$TOTAL_MISSING" -gt 0 || "$TOTAL_UNVERIFIED" -gt 0 ]]; then
+    echo "$ICON_SUCCESS Match:      $TOTAL_MATCH"
+    warn "Missing:    $TOTAL_MISSING"
+    [[ "$TOTAL_MISMATCH" -gt 0 ]] && echo "$ICON_ERROR   Mismatch:   $TOTAL_MISMATCH"
+    [[ "$TOTAL_UNVERIFIED" -gt 0 ]] && warn "Unverified: $TOTAL_UNVERIFIED (present, checksum timed out)"
 fi
 
 if [[ "$DRY_RUN" == true ]]; then
@@ -337,6 +371,8 @@ elif [[ "$TOTAL_MISMATCH" -gt 0 || "$TOTAL_MISSING" -gt 0 ]]; then
     echo "$ICON_ERROR Status: ISSUES FOUND — ${#SHARES_WITH_ISSUES[@]} share(s) need attention: ${SHARES_WITH_ISSUES[*]}"
     notify "Backup verify FAILED on $(hostname) → $REMOTE_SERVER_NAME — mismatches: $TOTAL_MISMATCH missing: $TOTAL_MISSING — shares: ${SHARES_WITH_ISSUES[*]}" \
         "Backup Verify" "warning"
+elif [[ "$TOTAL_UNVERIFIED" -gt 0 ]]; then
+    warn "Status: $TOTAL_MATCH verified, $TOTAL_UNVERIFIED could not be checksummed in time — NOT a clean run"
 else
     echo "$ICON_DONE Status: all $TOTAL_CHECKED files match across ${#VERIFY_SHARES[@]} shares ✅"
 fi