The predictive routing campaign keeps resetting dialing_strategy to progressive right after the apply step finishes. The 03:15 JST sync pipeline didn’t throw an error initially, but the UI shows the override vanished within minutes. Terraform 1.9.8 paired with nice-cxone 1.20.1. State backup confirms the config pushed correctly, yet the platform API quietly reverts it. Checked the nice_cxone_routing_predictive_outbound_campaign resource docs. Nothing mentions a backend validation rule blocking predictive mode on this account tier. Forced a terraform refresh and locked the concurrent_call_limit to a static value. The provider won’t retain the setting. API logs show a 422 on the PATCH request to /api/v2/routing/predictive-outbound-campaigns/{id}, but the CLI swallows the response code. Tried stripping down the block to just name and dialing_strategy. Same behavior. The drift detection script flags it immediately. Looking at the raw JSON payload, the agent_threshold field seems to trigger the rollback. Console output is doing jack all. The debug traces cut off right after the POST handshake.