Synthetic incident:
Observed: request queue grows while CPU remains normal
Question: is the bottleneck upstream bandwidth, connection handling or a slow dependency?
Immediate action: check existing metrics and contact the responsible provider if upstream capacity is affected
Exercise boundary: use recorded counters only; generate no flood traffic