Restarting can be useful when a project no longer fits its purpose or the existing structure prevents a necessary change. Preserve what the earlier attempt established before clearing the workspace.
Record the useful parts, known failures and unresolved assumptions. Identify the specific reason a fresh version would help. A new filename alone will not remove a problem you have not understood.
Keep one clear improvement
In a fictional guide redesign, the earlier version may contain a valuable worked example inside a confusing structure. Preserve the tested example while rebuilding the explanation around the reader's task. You do not need to discard useful evidence to justify beginning again.
Set a completion condition for the new attempt. Compare its performance on the original problem rather than judging it only by the relief of a fresh start.
If the same restart recurs, review the commitment itself. It may lack capacity, necessary knowledge or a convincing purpose. A responsible restart carries learning forward and changes the conditions that caused the earlier difficulty. Give the new version a real reason to produce a different result.