Prasanna Panchamukhi [off-list ref] wrote:

Our primary goal is to cap the maximum time taken by the GC to clean
up expired entries. We rely on user-space notifications to clean up
these entries from the hardware, so ensuring a predictable upper bound
is important for our use case.

Sure, but why can't we try to give a better default behavior?

while true; conntrack -L >/dev/null;done

basically does what you want already (but in a dumb way).

Regarding the adaptive strategy, we are using this sysctl to address
environments where the current average-based calculation delays the
cleanup of short-lived entries.

Yes, and I did propose to adapt the existing strategy to provide more
timely notifications.