Kaldi data directory validation warning policy for wav.scp and utt2spk

Kaldi data preparation relies on utils/validate_data.sh to check consistency of wav.scp, utt2spk and text files in a standard data directory.

The validation reports missing files and mismatched utterance IDs as warnings by default. The pipeline does not define a built-in constraint that promotes those warnings to fatal errors, so the decision to halt training or decoding rests with the recipe or operator.

With version-sensitive script defaults and custom recipes, the desired behavior for corpus verification is unclear.

Open questions

• Should missing entries in wav.scp or utt2spk be treated as fatal by default for training and decoding?
• What permission boundary should separate validation warnings from pipeline execution in a reproducible Kaldi workflow?