[bmc08gt]

When two users claim the same cash link at once, the loser sees "Something Went Wrong / Please try again later" and the claim is reported to Bugsnag. Retrying then shows "Cash Already Collected". The server confirmed why: the losing claim fails in its database transaction commit and returns gRPC INTERNAL (ErrorSubmitIntent.grpcStatus(13), "rpc server failure") instead of staleState(alreadyClaimed), so Session.receiveCashLink lands in its catch-all.

Bugsnag group 6a7b33b has 223 events from 37 users; 216 are on the claim (client.receiveCashLink) and 7 on createAccounts. This is item work from docs/specs/bugsnag-cleanup.md (code-payments/flipcash-client-orchestrator#91).

When the claim submit throws INTERNAL, CashLinkClaimRace.lostClaim re-fetches the gift card account info. If claimState is .claimed, Session throws staleState([], kinds: [.alreadyClaimed]), so the existing already-claimed catch handles it: info log, cashLinkClaims.record(collected: false), the "Cash Already Collected" dialog, and analytics with successful: false, with no ErrorReporting.captureError. Any other claim state, or a failed re-fetch, rethrows the original INTERNAL error into the catch-all unchanged. INTERNAL from createAccounts or elsewhere is untouched.

This is interim. Once the server returns STALE_STATE for the losing claim, the re-fetch can be removed.

Session takes a concrete Client, so receiveCashLink can't be driven in a unit test. The decision lives in CashLinkClaimRace, which takes the existing GiftCardAccountReading protocol, and Regression_6a7b33b covers it: claimed, not claimed, re-fetch throws, and non-INTERNAL errors (which never re-fetch).