Rails.error.report severity classification uncertainty in retry-prone deployments

Rails.application.config.active_support.error_reporter integrates with Rails.error.report and Rails.error.subscribe to emit exception signals with severity and context in Rails 7.0+. The goal is to establish a consistent severity-mapping and deduplication strategy that routes actionable exceptions to paging while suppressing noise from expected retries, deploys, and cascading failures. Severity taxonomy—distinguishing :critical from :error from :warning—remains application-specific, and no framework-default grouping key prevents repeat pages when the same error recurs across job retries, background requeues, or deploys. Context attachments such as controller actions, job classes, request IDs, and tenant IDs enable fingerprinting, but their stability and sanitization across process restarts introduce uncertainty in real-world grouping. Additionally, version differences between Rails 7.1 and 8.0 affect default subscriber behavior and severity defaults, requiring version-aware design choices.

• What severity threshold triggers paging for business exceptions versus framework errors in Rails 7.0+?
• Which class-name, request-ID, and context-key combination yields stable fingerprints across retries and deploys?
• How can context values be sanitized to prevent PII leakage while preserving grouping in Rails.error.subscribe handlers?