Most failed versions do not deserve permanent storage.
They deserve a decision.
Keep a version when it records why an attractive approach failed, contains a part worth reusing, diverges into a different viable work, or provides evidence needed to understand the finished result. Preserve enough context to make the failure legible.
Release duplicates that differ only by nervous saving. Delete exports that can be regenerated. Salvage material whose future use is credible and discard the rest through the appropriate waste stream. Photograph a bulky prototype when the image and measurements preserve everything the physical object can still teach.
The archive should not force every future search to pass through every prior
attempt. Name retained failures by the lesson or decision, not final-v7-old.
A short note can state what was tried, what happened, and whether the approach
should ever be reconsidered.
Failure has value after interpretation. Before that, it is volume.
The purpose of the archive is not to prove how hard the work was. It is to prevent useful evidence from disappearing and useless evidence from becoming another project to maintain.