Finch hashes from my browser:

c134752e-f3c7849
411b6d4e-3f4a17df
fe69e053-83ce3e87
d01ab0d3-2e200496
34a6bf44-ca7d8d80
da89714-4ad60575
9041608a-f23d1dea
8502ae4f-6b5e5ddf
1e528f0f-15305a2
6025934e-3f4a17df
c27fec31-c982d8da
7c1bc906-86bf56d9
47e5d3db-3d47f4f4
125b7f68-c7a5f76
41e765a5-3f4a17df
6557d030-f23d1dea
459a590c-ca7d8d80
3042ad4b-ca7d8d80
591576c8-f23d1dea
e56c5101-1410f10
5c72c65a-3d47f4f4
44827ee5-3f4a17df
334aa58d-3f4a17df
5485fc4d-3f4a17df
de47491b-3f4a17df
9773d3bd-ca7d8d80
93731dca-c037dd1c
41f007f9-f23d1dea
8fa604e0-ca7d8d80
c992f345-4ad60575
9e5c75f1-1bc64d8
f79cb77b-3f4a17df
fb4ab4bb-3d47f4f4
4ea303a6-3d47f4f4
bcc34a89-3f4a17df
6e6e0c7e-3f17a7d8
d92562a9-cfe3c2ea
7aa46da5-c946b150
2b33233e-1f7f5594
2c1d398c-3f4a17df
6973a1cf-3f4a17df
ad6d27cc-c6d02f41
da460ac8-f23d1dea
f3ea30a0-27a0c3c6
23496387-4ea78229
5a42b5d9-f23d1dea
344833e9-473e8c2e
3f273a97-25a103a9
4bc337ce-69465896
57789a80-141cae1e
9a2f4e5b-3066ab15
494d8760-52325d43
3ac60855-486e2a9c
f296190c-fc5c4256
4442aae2-e1cc0f14
ed1d377-e1cc0f14
12e17bc5-e1cc0f14
75f0f0a0-e1cc0f14
e2b18481-6e597ede
e7e71889-e1cc0f14
3a4029d-f23d1dea
bbb8f811-f23d1dea
6a51bb09-f23d1dea
cc73f8a1-a3a14831
8834fcca-f23d1dea
530efe1f-3f4a17df
3f33c9bd-ca7d8d80
81c6897f-3d47f4f4

What it sounds like.
Attached .wav-file.

Deleted: example_reverb.wav 1.5 MB

example_reverb.wav 1.5 MB Download

How often does this happen?
Can you narrow down the version a bit? I.e., does it work fine in 67 but not in 68?
From the hashes, it looks like there's at least one WebRTC experiment active (a gain controller one), could that be the culprit?

Is example_reverb.wav the debug recording of the output? If not, can you do one and upload? Also, the AEC dumps would be good to compare with.

Gustaf, assigning this to you for Oskar and Fredrik's questions, we'll take over investigation afterwards.

We managed to reproduce the bug today.

This time I was using a Linux laptop (Lenovo X1 Carbon) with Chrome 68.0.3440.7 (Official Build) dev (64-bit).

I have only seen this issue with Chrome 68.

Here are recording from both ends:

reverb_bug.aecdump is from the side where the issue is heard.
reverb_bug_farend.aecdump is the other end.

https://drive.google.com/drive/folders/1Lf1p33HWLyuFpmNeu3LSkxgKvqXsNzLR?usp=sharing

Looks like the audio going out from APM on the farend is ok.
The audio coming in to APM on the nearend has the reverb issue.

Posting my current hashes. Don't know if they are of any value to you.

c134752e-6d5b49a5
411b6d4e-3f4a17df
fe69e053-94941f92
d01ab0d3-2e200496
e202a358-3f4a17df
34a6bf44-f23d1dea
a6674cf-f22a02fa
da89714-4ad60575
9041608a-3f4a17df
8502ae4f-f23d1dea
1e528f0f-ca7d8d80
afb5d7b8-a2bfa3a0
ca05d627-3f4a17df
7c1bc906-6790560b
47e5d3db-3d47f4f4
125b7f68-65bced95
6557d030-6557d030
4dc30737-b8a5ea08
3042ad4b-f23d1dea
591576c8-a8bffb1
267255c3-f4950e99
44827ee5-43146c13
5485fc4d-3f4a17df
937cad47-65bced95
9773d3bd-ca7d8d80
93731dca-e89d496c
41f007f9-f23d1dea
8fa604e0-a8bffb1
c992f345-4ad60575
9e5c75f1-e406a769
350fabdd-34b13816
f79cb77b-3f4a17df
fb4ab4bb-3f4a17df
4ea303a6-9f93d4fc
bcc34a89-3f4a17df
6e6e0c7e-f23d1dea
d92562a9-65bced95
2b33233e-1f7f5594
2c1d398c-3f4a17df
6973a1cf-f23d1dea
58a025e3-c2b41702
4932440-d21eb72d
f242806f-5810b593
da460ac8-65bced95
5a42b5d9-ca7d8d80
344833e9-1525b35b
3f273a97-25a103a9
4bc337ce-69465896
57789a80-ca7d8d80
9a2f4e5b-f46d6bfd
494d8760-52325d43
3ac60855-3ec2a267
f296190c-9e248081
4442aae2-e1cc0f14
ed1d377-e1cc0f14
12e17bc5-e1cc0f14
75f0f0a0-a5822863
e2b18481-6e3b1976
e7e71889-e1cc0f14
3a4029d-3a4029d
bbb8f811-3f4a17df
6a51bb09-f23d1dea
41aa6aaa-2b71c400
cc73f8a1-a3a14831
8834fcca-cf4f6ead
530efe1f-3d47f4f4
3f33c9bd-ca7d8d80
81c6897f-3d47f4f4