Skip to content
Notifications
Clear all

Anyone else having issues with Zscaler and MS Teams media streams dropping?

2 Posts
2 Users
0 Reactions
4 Views
(@Anonymous 447)
Joined: 1 week ago
Posts: 17
Topic starter   [#316]

Hi everyone! First time posting here, and honestly a bit nervous to jump in. 😅 I've been tasked with evaluating how various AI coding assistants perform under different network conditions, which is how I ended up deep in our corporate Zscaler setup.

Lately, I've been running into really inconsistent behavior specifically with Microsoft Teams calls. The audio/video streams will just... drop. The call doesn't necessarily disconnect, but the media freezes or goes silent for 5-10 seconds. It happens a few times per hour-long meeting.

My setup/testing context:
* Our org uses Zscaler ZIA (Internet Access) with the client connector installed.
* Teams is set to bypass the proxy for media (UDP 3478-3481, etc.), as per Microsoft's recommendations.
* I'm not on VPN, as we're a Zscaler-first shop.

The weird part is it *seems* tied to when my evaluation scripts are pulling data from cloud AI APIs (like OpenAI or Anthropic). It's like the increased concurrent connections from my tools somehow affect the Teams media path through Zscaler.

Has anyone else experienced this? I'm trying to figure out:
* Is this a known Zscaler/Teams tuning issue?
* Could it be related to specific Zscaler policies or bandwidth profiles?
* Are there client connector logs I should be checking first?

I feel like I'm troubleshooting two things at onceβ€”my evaluation project and now the network! Any pointers would be so appreciated.



   
Quote
(@martech_tester)
Trusted Member
Joined: 3 months ago
Posts: 32
 

Oh man, this feels familiar, but from the other side of the stack. We saw something like this when our sales team was running heavy demo loads through the same ZIA policy. Their screen-sharing would glitch out.

> *seems* tied to when my evaluation scripts are pulling data from cloud AI APIs

That's a big clue. Zscaler can sometimes treat those high-volume, persistent API connections differently, maybe even flagging them. It could be creating temporary contention or even a micro-burst that briefly chokes the UDP media bypass. Have you checked if your scripts are forcing TCP sessions that might be getting inspected? That's where we'd see the hiccup.



   
ReplyQuote