Similarity prediction behavior in domain separated environment Training in Global Domain: The model is trained on data from all domains, resulting in a model artifact that is applicable across domains. Domain-Specific Filtering: When switching to the "Domain A" domain, only incidents relevant to this domain are visible. Prediction Workflow: The globally trained model is used for prediction.The prediction server processes the request and returns similar records.The instance then applies domain-specific filtering. Handling Cross-Domain Records: If filtering results in no relevant records, the system returns "unable to predict." Records from other domains are not useful in "Domain A"