Production readiness

Production readiness — errors, retries, state, security, and observability

Connection & state

• Provider detection order: window.$onekey.ethereum → OneKey in multi‑provider list → window.ethereum

• Ask for connection on user intent (avoid unsolicited popups)

• Listen to accountsChanged/chainChanged and recreate clients as needed

Errors & retries

• Standard codes: 4001 (rejected), 4902 (unknown chain)

• Network/node errors: exponential backoff; capture and classify for telemetry

• Tx polling: set timeouts and user guidance to avoid dead‑ends

Security & permissions

• Nonce‑based sign‑in to prevent replay (see “Authenticate users”)

• Don’t trust frontend‑supplied chainId/address server‑side; re‑validate

• Limit dangerous RPCs; expose only what you need

Performance & UX

• Cache read‑only data sensibly; prefer Multicall for many reads

• Show clear network/account and fee information in UI

• Prepare deeplinks and Universal Link fallbacks for mobile

Observability & support

• Instrument key paths: connect rate, rejection rate, tx success rate

• Record error codes with context (chain/node/method)

• Provide copy‑ready error details for support and debugging