stable-rc optimizations, less often builds or build on trigger?

It seems our trigger are being often caught up in the middle of stable-rc updates, they might update tree during more than 1 hour, and we often unnecessarily retrigger the build. As i remember, there was discussion that it is not necessary to test tree on every change, but only on some email (?). This will help to reduce load on build cluster:

Start by tracing the stable-rc update trigger and how it starts builds while the tree is being updated. Determine which trigger policy is intended, then validate that redundant builds are avoided without skipping the required stable-rc testing and that build-cluster load is reduced.

Written by the indexing model from the issue text.