dataset_identity_index_unavailable
The upload succeeded but its identity index could not be built.
dataset_identity_index_unavailable appears as a stored failure on a dataset upload, not as an HTTP error envelope. Read it from the upload resource rather than from a response status. The uploaded file arrived intact; building the index that makes its rows addressable did not finish.
The failure is marked retryable: true, so the upload can be processed again without re-sending the file.
Recovery guidance
Retry upload processing. The stored source is still present, so this does not require a new upload and will not create a second dataset.
If retries keep landing on the same failure, the upload is not the problem — preserve the upload ID and the requestId from the retry response and escalate rather than looping. Do not fall back to creating a fresh upload as a workaround: that leaves the original upload stuck and adds a duplicate dataset alongside it.