What's the issue?
When performing a restore from S3, we create a plan operation to download (and cleanup) Rift indexes. For restores which have the bucket data step disabled, we're downloading these files and not using them.
What's the fix?
We shouldn't create this operation when the data step is disabled. Note that we already have some logic which skips the creation when no data is being transferred (e.g. due to filtered collections).