The Zombie Call Bug
# support
f
Hey everyone, we’ve run into a nasty edge case that I’ve started calling the Zombie Call Bug – like with anonym numbers, but thats a different case, i solved. What happens? During a warm transfer (warm-transfer-experimental) the call suddenly loses ownership: - Vapi no longer controls the session. - Twilio shows the call as “in progress” but with no proper signaling or media. - Neither system is able to trigger fallbacks or a disconnect. The result - The caller just hears nothing. - No fallback is played, no transfer completes. - The call stays alive as a “zombie” leg for up to 10 minutes until Vapi’s hard timeout kills it. Why it’s critical It’s not just a UX issue – it blocks lines, burns minutes, and leaves customers confused because they don’t get any feedback. Since neither Vapi nor Twilio claims ownership, you’re left with an orphaned call session in limbo. What we need Has anyone else hit this edge case or found a reliable workaround? Right now I’m considering building an external watchdog (kill switch via API after ~45s of silence), but I’d love to know if the team is already working on a fix for transfers losing session state. Thanks 🙏 Call-ID: 7e6439a2-c6ba-4ef7-a133-df7528d00272 Assistent-ID: d9d5bb69-e45d-4973-9c84-fed06d104a04 – Felix https://cdn.discordapp.com/attachments/1418853227549692016/1418853227708944504/image.png?ex=68cfa18a&is=68ce500a&hm=21fbe8054cdafc221e79f47dd724cf2c17214df385e76b5cdc15ed853dcbabb5&
It happens since last friday, the 12.09.25. Maybe its a Twilio thing, but i cant look into the logs deep enough.
v
Suggestion: You can mitigate the "Zombie Call Bug" by configuring your Transfer Assistant to automatically end idle calls after a set period of silence using the
silenceTimeoutSeconds
parameter. For example:
Copy code
json
{
  "transferPlan": {
    "mode": "warm-transfer-experimental",
    "silenceTimeoutSeconds": 45
  }
}
This will ensure any call with no media or signaling for 45 seconds is programmatically ended, preventing orphaned sessions and improving user experience. Reference: [August 28, 2025 Changelog](https://docs.vapi.ai/changelog/2025/8/28) Source: - [August 28, 2025 Changelog](https://docs.vapi.ai/changelog/2025/8/28)
f
It's not working with silence timeout, it looses connection to VAPI, like i said. I tried it all, to fix it with an 10+ hours and help from different people that also had no solution, but to switch to retell.
This is a major incident, and can't be worked aroun. Voicemail detection is the only think that works at the moment and gets the transferCancel tool triggered.
We need a maxDuration for the transfer in the state ringing in my opinion.
c
Thanks for the call ids. We haven't really seen this issue much before but you gave some pretty good detail about it. We will review it and get back to you.
h
hi, can I send you pm, I will appreciate it!
any update?
f
Hey kyle, please call me, if necessary. I have huge interest in fixing it. Since two weeks i have a lot of expenses through this issue.
sent you a pm
@Herbynator @User @User @Kyle During a warm transfer, the call keeps ringing. After X seconds, the PBX sends a busy signal. From that point, the transfer fails – but instead of ending cleanly, it turns into a zombie call.
i switched to retell now.
c
Sounds good.
2 Views