Lab / / experiment
A control that proves the faults landed
"Every recoverable packet came back" is a claim a broken injector could also make. Running the same seeds with retransmission off left 19 / 113 / 314 permanent holes.
Controls, WebRTC, Measurement methodologyKeryx- Machine
- UNKNOWN — the integration tests do not state one
- Commit
- `cacb70c`, 2026-08-25
- Stream
- 2,098 packets
- Permanent holes, retransmission off
- 19 / 113 / 314 at 1% / 5% / 15% loss
- Repairs, retransmission on
- every packet the receiver could detect as missing
The fault injector splices in under the sender's SRTP and drops, duplicates, reorders and delays media datagrams while STUN, DTLS and RTCP pass through untouched. That placement is the design: injecting loss below SRTP makes the fault indistinguishable from a real network to everything above it, and injecting only into media datagrams keeps the handshake healthy, so the test measures the media path rather than whether the connection survived.
At 1%, 5% and 15% uniform loss and under ten-packet bursts, every packet the receiver could detect as missing came back. That is the claim. It is also a claim a completely broken injector could satisfy, because an injector that injects nothing loses nothing.
So I ran it again with retransmission off
Same seeds, retransmission switched off: exactly those packets stay lost. On a 2,098-packet stream that is 19 / 113 / 314 permanent holes, monotonically increasing with the loss rate, as it must.
| Arm | Result |
|---|---|
| retransmission on | every detectable missing packet returned |
| retransmission off, 1% / 5% / 15% | 19 / 113 / 314 permanent holes |
| stream size | 2,098 packets |