[currently more of a placeholder than a task; sub-tasks to follow]

There exists all of:

• a lack of a strong understanding of what the ratio between edge egress traffic and transport backhaul "should" be, under conditions ranging from warm-cache steady-state, to bringing up a 'cold' edge DC, to various attack scenarios
• a lack of an intuitive awareness of & observability of as to what some of the backhaul traffic is, as was witnessed in T263206: cr1-codfw<->cr1-eqiad link saturation
• several possibilities to optimize said ratio -- most past caching improvement efforts have focused specifically on per-request hitrate, and not on backhaul byte conservation.