[bmc08gt]

Bugsnag group 6a7b33b (receiveCashLink, 411 events) holds eight error types. Catch-all captures that pass no id: group on #fileID:#function only, so every error thrown in one function lands in one group. This is Decision 4 (iOS half) in docs/specs/bugsnag-cleanup.md, from the triage brief .claude/plans/2026-10-30-bugsnag-6a7b33b.md on docs/bugsnag-cleanup-triage.

The fold is in ErrorReporting.capture, not at the call sites. A new ErrorReporting.groupingHash appends the error's type and value when id: is nil, so every existing catch-all gets split without editing each one. An explicit id: still replaces it, so sites that already pick their own id keep their current groups.

Associated values do split groups: grpcStatus(13) and grpcStatus(14) get different hashes. ErrorSubmitIntent prints the status code, and INTERNAL and UNAVAILABLE are different problems. This matches the event title, which already uses "\(error)". The cost is that an error whose description embeds free-form text would fragment into one group per message. I did not audit every error type for that.

Severities are unchanged: ErrorFetchBalance.notFound stays .info and transient RPC errors stay suppressed. Existing catch-all groups will start fresh on the first release with this change, so old groups need closing as superseded.

ErrorReportingGroupingTests covers different types, same error, differing associated values, different functions, and explicit id:.