(Windows) sf.write runs into LibsndfileError: System error

[simonriou]

sf.read(filepath) works as intended, it successfully finds the file. However, when trying to write to the file (replace its content), this error triggers

File "C:\Users\User\Documents\someFolder\src\util\files.py", line 92, in set_fixed_duration
    sf.write(file_path, audio_trunc, SAMPLE_RATE, format='FLAC')
  File "C:\Users\User\miniforge3\envs\speech\lib\site-packages\soundfile.py", line 363, in write
    with SoundFile(file, 'w', samplerate, channels,
  File "C:\Users\User\miniforge3\envs\speech\lib\site-packages\soundfile.py", line 690, in __init__
    self._file = self._open(file, mode_int, closefd)
  File "C:\Users\User\miniforge3\envs\speech\lib\site-packages\soundfile.py", line 1265, in _open
    raise LibsndfileError(err, prefix="Error opening {0!r}: ".format(self.name))
soundfile.LibsndfileError: Error opening 'data/train/speech/103-1240-0000.flac': System error.

This is my code

def set_fixed_duration(directory):
    """
    Set all audio files in the specified directory to a fixed duration by truncating or padding with silence.
    Replace the original files with the adjusted ones.
    """

    target_len = int(SAMPLE_RATE * DURATION)

    for filename in tqdm(os.listdir(directory)):
        if filename.endswith('.flac'):
            file_path = os.path.join(directory, filename)

            # Load the audio file
            audio, sr = sf.read(file_path)
            if sr != SAMPLE_RATE:
                raise ValueError(f"When setting duration | Expected sample rate {SAMPLE_RATE}, but got {sr}")
            
            # Convert to mono if needed
            if audio.ndim > 1:
                print(f"Converting stereo to mono for file: {filename}")
                audio = np.mean(audio, axis=1)

            # Print first 10 samples for debugging
            # print(f"Original audio samples (first 10) for {filename}: {audio[:10]}")
            
            L = len(audio)

            # Adjust duration
            if L > target_len:
                # Truncate
                audio_trunc = audio[:target_len]
            else:
                # Pad with silence (zeros)
                pad_width = target_len - L
                audio_trunc = np.pad(audio, (0, pad_width), mode='constant', constant_values=EPSILON)

            # Replace the original file with the adjusted one
            sf.write(file_path, audio_trunc, SAMPLE_RATE, format='FLAC')

I have soundfile v0.13.1 and libsndfile v.1.2.2
I am on windows 11