Troubleshooting

Concrete fixes for the failures you are most likely to hit. Pair each with a Trace log and quality snapshots — see Diagnostics.

Registration fails

401/407 loop — wrong username/password/realm. The SDK does a bounded stale-nonce retry; a persistent challenge means credentials. For 3CX, use the auth ID, not the extension number (3CX).
No response to REGISTER — registrar host/port unreachable, or the box expects TLS. Check SipServer and network path.
Binding lingers after restart — fixed: unregister reuses the registration Call-ID + CSeq (RFC 3261 §10.2.2). Prefer an awaited UnregisterAsync on shutdown over relying on dispose.

SDP advertises a wrong/unreachable media address → check c=/m= in the trace. The SDK advertises a routable (non-loopback) address; on complex NAT you may still need an SBC. See NAT and SIP trunks.
No common codec → align PreferredAudioCodecs with the peer; keep G.711 (PCMU/PCMA) as a baseline.

High jitter/loss in the quality snapshots → a network problem, not the SDK. The adaptive jitter buffer compensates within limits.
A stalled RTP timestamp burst (comfort noise / repeats) no longer derails the jitter estimator (fixed in 4.3.1).

Confirm RFC 4733 telephone-events are negotiated in the SDP and the peer's DTMF mode matches. Inbound DTMF fires DtmfReceived — remember it can arrive on two threads (Events).

With SrtpPolicy.Required, a peer that can't do SDES SRTP yields a failed call by design (no silent downgrade). Confirm a=crypto in the trace and the peer's SRTP setting. Use Optional to allow fallback. See SRTP/SRTCP.

Explicitly HangupAsync active calls and UnregisterAsync lines before Dispose(); the dispose-time cleanup is best-effort (Lifecycle & dispose).

Capture a Trace log of the failing call and open a conversation at info@bechstein.digital.