The algorithm we use to store releases files works well with concurrency, where separate releases might be trying to store the same files, but if an import fails midway for some other reason you end up with a partially imported tarball which isn't easily fixable.
One solution might be to add a tarball_status field to releases which can be used to recover an import if it didn't successfully fully import.
The algorithm we use to store releases files works well with concurrency, where separate releases might be trying to store the same files, but if an import fails midway for some other reason you end up with a partially imported tarball which isn't easily fixable.
One solution might be to add a
tarball_statusfield to releases which can be used to recover an import if it didn't successfully fully import.