On Wed, Jul 03, 2019 at 02:10:05PM -0400, Sasha Levin wrote:

> Sure, but I don't see any work upstream on trying to correct this?

There's nothing to do to correct it, it's a normal thing that
might happen in an audio system.  Probably users on the old
kernel who are bothered by it either did a local kernel change or
inserted a delay at the application level.  Correcting it again
separately will double the delay, the problem is changing it
affecting system integrations.

> These sort of things are a reason why users stick with the same kernel
> for years, which is what we'd like to avoid.

Domain specific knowledge can be helpful...