A drifted owned search index waits through the 60 s backend retry #1368

Problem

When service materialization finds an index that this content owns but whose Cortex relations mapping fails the product contract, it fails with reason native-product-contract-mismatch (src/aptl/core/deployment/_service_index_materialization.py:362, dev 3bdccc7). Materialization never recreates an owned index (test_owned_existing_index_is_idempotent_no_recreate), so a second apply of the same plan reads the same mapping and fails the same way.

The provisioner still files that failure under aptl.provisioner.backend-start-failed, the only code in _RETRYABLE_APPLY_DIAGNOSTIC_CODES (src/aptl/backends/raes.py:96). For a pack whose adapter admits BEFORE_BACKEND_RETRY, aptl lab start therefore waits 60 s, runs the repair hook and applies again before it reports the refusal (src/aptl/core/lab.py:2542-2563).

The fix for #1325 ends the start without the wait and the retry for unowned-collision only, through a structured LabResult.unretryable flag that the compose step sets for the reasons in _UNRETRYABLE_REASONS (src/aptl/core/deployment/_compose_service_index_realization.py). #1281 (for #1280) adds the drifted-index message and keeps the retryable code for this reason. Neither change covers it.

Scope

• Add native-product-contract-mismatch to the reasons that a retry of the same admitted plan cannot clear, so the provisioner reports it with aptl.provisioner.backend-start-refused and skips the wait and the retry.
• Keep the retry for transient start failures and for readiness timeouts (native-query-failed).
• Keep the message and hint text that fix(deployment): explain how to recover from a retained index whose mapping drifted #1281 adds.

Done when

• A native-product-contract-mismatch failure ends aptl lab start without the 60 s wait and the retry, even when the adapter admits BEFORE_BACKEND_RETRY.
• Transient backend start failures still get the single retry.
• Tests cover both outcomes.

Related

• An unowned index collision waits through the 60 s backend retry #1325; Explain how to recover from a retained service index whose mapping drifted #1280 / fix(deployment): explain how to recover from a retained index whose mapping drifted #1281 (drifted mapping of an owned index); fix(deployment): explain how to recover from an unowned service index #1264 (for Explain how to recover from an unowned service index #990).