Precision of statistic data

Dear All,
Recently I try to test the adaptive feature of jitsi using the Netlimiter (throttering the bandwidth to simulate the fluctuation). I can see the deviation between the data shown on the Jitsi-meet and those from the Netlimiter as in the enclosed picture.

Would you please confirm that at least we are quite confident on the data from Jitsi-meet.
Another question is that when the bandwidth is released back to a high value (high bandwidth) the video on jitsi-meet (it was turned off previously to save bandwidth) is not turned on again automatically. I have to refresh by myself to see the webcam again. How can I activate / fix so that jitsi-meet is more responsible to such change in the bandwidth quality?

Many thanks for your helps

Did you gave it time to return to normal?

Dear @damencho, I did wait a while. But statistic data in both GUI do not seem to converge.
And the jitsi-meet , once it was turned off for saving the bandwidth, doesn,t seem to recover the webcam. Can you give me some hint to debug the problem please
Many thanks

So there are few things here, when this happens does chrome starts sending video maybe check in chrome://webrtc-internals.
Second if that is an issue it should be in the bridge, so then make sure you test with the new bridge(jitsi-videobridge2) and it will be debugging in the new bridge.
Do you see same behaviour with jvb2?

Dear @damencho
I use the code checked out last week and I didn,t observe there is a jvb2. I will check the code from the nightly built
Many thanks

To install jvb2, you need to be using unstable debian repo and you need to do:

apt-get remove jitsi-meet
apt-get install jitsi-videobridge2

Thank you @damencho. But I really want to take a look into the algorithms in bridges 2. IS the source code available ? Also can I see the new features of this version.
Best regards

The source code is in master in the old bridge is in a branch jvb_1.0