Do you want this feature just for layer-7 HTTP request/response, or for all of the layer-4 tunnel connections?

Layer 4, please. I do love having socks5, it's so much more efficient. In my tests it consumes way less data for the same URL compared to http proxy.

Design note: please do not log the total bytes received on top of chunks' logs.

Example if downloading a 1GB file and my speed was 1MB/s:

1. log received 1MB
2. log received 1MB
3. ...
4. do not log 1GB received

Thank you Mr. Zhang!!!

For tunnel connections, we don't known whether it's downloading files. What can be done is that, log when reading/writing to client connections at some nearly fixed time interval.

That works!

Note: auth and its refresh interval must be respected. As in, a user could be valid now, but invalid at a later chunk/interval. In that case the connection must be terminated.

Should this be a separate issue? Is this already implemented? (I haven't tested.)

Note: auth and its refresh interval must be respected. As in, a user could be valid now, but invalid at a later chunk/interval. In that case the connection must be terminated.

Should this be a separate issue? Is this already implemented? (I haven't tested.)

You can set a user to blocked state and then it will be blocked after idle_check_interval (5min by default).

The enovy way: https://www.envoyproxy.io/docs/envoy/latest/intro/arch_overview/observability/access_logging#periodic-access-logs.
The plan is to add a task_log_flush_interval and task_log_on_stage_change config options in server config, and just do task log at that flush interval or when task stage changed (new -> connected -> relay).
@NinoSkopac does this work for you?

Sounds good.

I guess for my particular use case I only need to use task_log_flush_interval?