The problem
The existing frontend would silently fail whenever the primary RPC returned a transient 502, display stale quotes during reorgs, and had no way to diagnose failed transactions after the fact. Support tickets were the primary monitoring system.
What we built
A resilient frontend with an RPC fallback layer, reorg-aware data fetching, and a client-side telemetry pipeline that captures every wallet interaction into a structured event stream.
Highlights
- Multi-provider RPC router with health checks and circuit breakers
- Reorg detection in the data layer; quotes invalidated and re-fetched on reorgs
- Frontend OpenTelemetry pipeline with wallet-anonymized session traces
- Transaction replay tool for support, built on the telemetry stream
Outcome
- Failed-interaction rate dropped from 6.2% to 0.4%
- Support ticket volume dropped 70% quarter-over-quarter
- Mean time to diagnose a failed tx dropped from hours to minutes
Stack
TypeScript · Viem · Wagmi · TanStack Query · OpenTelemetry · Vercel