[planckscnst]

Yes, the cost/latency trade-off is exactly why this can be interesting. Significantly fewer or less-impactful outliers can save a lot of money (or allow a better SLA with the same cost).