Problem
The /getTranscript endpoint currently uses glob.glob and max(key=os.path.getctime) to identify the downloaded subtitle file.

The Logic Flaw
This assumes that the file currently being processed is always the newest file in the subtitles/ directory.
In a concurrent environment (multiple users), if User B's download finishes slightly after User A's download, User A's request thread will incorrectly identify User B's file as the "latest" and return the wrong transcript.

Impact

1. Data Leak: Users receive transcripts for videos they did not request.
2. Crash: If the "latest" file is deleted by another thread before it can be read, the server throws a FileNotFoundError.

Proposed Solution
Since the filename is deterministic (subtitles/{video_id}.vtt), we should explicitly construct the file path using the video_id instead of searching the directory.
I am working on a fix for this and will submit a PR shortly.