Picard currently always tries to guess the file format based on file content. This is very IO expensive and in many cases not needed.

Change the logic to:

• Try guessing the format by file extension first
• If no extension or no matching file extension is available, guess by file content
• If format was guessed by extension and loading the file with mutagen fails with an exception, retry with guessing by file content. Retry loading if the newly guessed format is different from the format guessed by extension.

See discussion at PICARD-975