per_server.connection.match 'ignores' parent model

A primary use of per_server.connection.max is to limit the number of upstream connections to an origin.

per_server.connection.match allows you to hash against {fqdn,ip} to apply those limits. Such that many hostnames going to a single address can have a separate limit defined.

However: In a multi-tiered deployment where "parents" are defined, this is seemingly not possible. The fqdn seems to come from current.server->name which in turn comes from the parent_info struct, and as such fqdn:ip is 1:1 and 'both' no longer has any value

child ->
child -> parent -> originA
child -> parent -> originB
child ->

Only being able to apply a meaningful fqn:ip limit outbound from the parent doesn't allow inbound to be collapsed. So a slow originA will cause increased impact on originB.