fix(watch): don't backfill existing archives at watch startup

Co-authored-by: Belphemur <197810+Belphemur@users.noreply.github.com>

README.md

@@ -78,6 +78,14 @@ Watch a folder for new CBZ/CBR files and optimize them automatically:
 cbzconverter watch [folder] --quality 85 --override --format webp --split
 ```
 
+Watch mode only reacts to filesystem events that occur *after* it starts; it does not
+scan and optimize files that already exist in the folder when it starts. Run the
+`optimize` command first if you need to process an existing library, then use `watch`
+to keep it up to date going forward. The only exception is a directory that gets
+created/moved into the watched tree while watch mode is running: since no per-file
+event is emitted for files already inside it, its existing archives are processed once
+when the directory is first detected.
+
 Or with Docker:
 
 ```sh

cmd/cbzoptimizer/commands/watch_command.go

@@ -100,10 +100,16 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 	debouncer := newEventDebouncer(debounceDelay, queue.Enqueue)
 	defer debouncer.Stop()
 
+	// Note: existing archives already present under path when the watch
+	// starts are intentionally left untouched. Watch mode only reacts to
+	// filesystem events going forward; use the `optimize` command to process
+	// a library's existing contents. Archives inside a directory that is
+	// created/moved into the watched tree *after* startup are still
+	// back-filled below, since only the directory itself generates an
+	// fsnotify event.
 	if err := addRecursiveWatch(watcher, path); err != nil {
 		return fmt.Errorf("failed to watch path %s: %w", path, err)
 	}
-	backfillExistingArchives(path, debouncer.Trigger)
 
 	for {
 		select {