The problem seems to be that in the second recording,  chameleon records no data.

I can not reproduce this symptom by calling
>>>p.StartCapturingAudio(3)
>>>p.StopCapturingAudio(3)
repeatedly using test_server script.
The recording always works fine.

I think there must be something wrong in the test.

Thanks to Chihchung's help, the root cause is that receiver thinks audio data is broken. This is dependent on Cros side behavior.
For Chameleon side, we just need to reset audio logic on receiver whenever we start recording.
The fix is posted.
https://chromium-review.googlesource.com/chromiumos/platform/chameleon
I will do more testing on other boards.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/chameleon/+/67c0e6ab71686879b17b733974ceaefc1daa53d7

commit 67c0e6ab71686879b17b733974ceaefc1daa53d7
Author: Cheng-Yi Chiang <cychiang@chromium.org>
Date: Mon Jun 27 10:50:31 2016

input_flow: Resets audio logic before capturing audio data

For some ChromeOS boards, receiver on Chameleon treats audio data stop as
an error. When receiver is in error state, it does not dump data
anymore. Reset audio logic before recording so receiver can start
dumping new data.

BUG=chromium:623446
TEST=run audio_AudioVolume.hdmi test on nyan_big.

Change-Id: I7fc04714adb6c4e05208906206e1f600a99122d7
Reviewed-on: https://chromium-review.googlesource.com/356336
Commit-Ready: Cheng-Yi Chiang <cychiang@chromium.org>
Tested-by: Cheng-Yi Chiang <cychiang@chromium.org>
Reviewed-by: Shyh-In Hwang <josephsih@chromium.org>

[modify] https://crrev.com/67c0e6ab71686879b17b733974ceaefc1daa53d7/chameleond/utils/input_flow.py
[modify] https://crrev.com/67c0e6ab71686879b17b733974ceaefc1daa53d7/chameleond/utils/rx.py

You started fixing this bug over two years ago. Are you still working on it?