A production WAF on Front Door Standard is mostly about rule budget and false-positive control, not stacking every managed set on day one. Cap custom rules to the paths that actually see abuse, pin exclusions to Host plus path, and keep bot or rate policies separate from signature rules so a single noisy pattern cannot burn the whole custom-rule ceiling. Prefer blocking at the edge for known bad paths and logging-only for new managed rules until you measure client impact.
Ops prove-out after enabling or tightening a policy: replay a clean canary Host and a known-bad path through the same front door and confirm only the bad path gets challenged or blocked. Watch challenge and 4xx rates by Host for one deploy window — a spike on a previously quiet Host usually means an exclusion or match condition drifted. Roll the policy under peak and compare origin request rate; if origin climbs while the edge still reports allow, you are logging instead of enforcing where you think you are.