Allow running Front Door (WAF) in a learning mode
If should be possible for Front Door to learn what requests look like for a site, and then white-list those request patterns, and at the same time indicate that requests that don't match the learned pattern should be considered more likely to be "bad requests".
There should be a GUI where blocked requests are listed, where you could select if it's an OK request or not.
Also, Front Door should know" when there's a new deployment of the backend, and then automatically enter a a semi-learning mode for a period.
During this period FD should be less prone to block new request patterns (NOTE: new, as in request patterns not previously seen, especially for completely new paths).
Not everyone use CD on the production environment
Tomáš Maleček commented
When practicing CD, Front Door would be permanently in the semi-learning mode.
There would have to be an interface for notifying Front Door about deployments.
Do not see a way to implement this in any reasonably simple-to-use way.