Telenyx webphone

The job in one sentence: Our agents take inbound sales calls in the browser via a Telnyx WebRTC softphone, and calls are dropping/failing . We need someone who has shipped production browser-telephony before to root-cause it, fix it, and make it provably reliable. Stack: Next.js 15 (App Router) + TypeScript, deployed on Vercel @telnyx/webrtc (browser SDK) for the softphone Telnyx TeXML voice app + SIP; calls originate from a third-party dialer (CallGrid, which is Twilio under the hood) → Telnyx → browser Postgres (Neon) + Drizzle ORM; Inngest for async/billing jobs Opus/G.722 codecs, getUserMedia, RTCPeerConnection What's actually broken (be honest with yourself about whether you can fix these): Inbound calls intermittently fail to bridge: ring → answer → dead air or immediate hangup WebRTC reconnect/registration churn (telnyx.error storms, overlapping registrations, websocket flap) getUserMedia / mic-device edge cases dropping calls silently First-few-seconds audio quality (APM/codec/jitter) on some calls We lack clean server-side observability into why a given call ended (SIP hangup cause/source, caller vs callee) Billing integrity tied to call lifecycle (bridged/duration/disposition) must stay correct