Recovering a Nimble Package After a Failed Upgrade

When nimble upgrade is interrupted by a network failure or other error, the affected package directory can be left with partial or corrupted files. Nimble performs in-place replacement without a transactional rollback or automatic backup of the previous version, so the package may become unusable. The nimble check command verifies integrity but does not repair or restore the package. Error messages and the exact failure behavior vary across nimble versions, making recovery steps difficult to standardize.

Given these constraints, what is the expected manual recovery workflow to return a package to a consistent state? Is there a reliable way to identify that a package is partially upgraded, independent of version-specific error strings? And are there any documented community approaches for achieving atomic upgrades, such as staging a new version and renaming directories?