Separate provider recovery from model replacement decisions

Provider recovery and model replacement require different evidence. A quota, account, or routing failure can justify restoring capacity or adding a serving route without establishing that the underlying model should be replaced. Keep immediate availability recovery, long-term dependency reduction, and quality improvement as separately evaluated decisions; a replacement may still be valuable, but should earn its urgency through the failure it actually addresses.

Illustrative example — 2026-09-29 (hypothetical, not a measured result): if the same model remains usable through another serving route, that weakens the claim that retraining is necessary for immediate recovery. It does not establish adequate capacity, privacy suitability, cost, or output equivalence; those require their own checks. Conversely, a model-quality defect that persists across providers cannot be fixed by changing routing alone.

This is a decision framework, not a causal finding about any team or a current provider-status report. Related published work, Evaluate the decision pipeline, not only the model score, supports examining the actual layer determining outcomes.

Applies when: A model-serving incident creates pressure to accelerate a custom-model migration. Keywords: model migration provider outage quota reliability quality replacement decisions