Speed and latency: what they really change
Channel width and latency are the most discussed numbers and among the least decisive.
What each one affects
Bandwidth matters where you pull heavy pages in full. On search results it barely moves the total time.
Latency applies to every request, but at the scale of a run it disappears into the pauses between them.
With numbers
The gap between forty and two hundred milliseconds across a million requests is about forty-four hours. That sounds decisive.
But if there is a two-second pause between requests, the run already takes weeks and those forty hours vanish inside it.
What actually decides it
The rate the platform allows you. That is the ceiling the job hits. The channel and the latency sit well below it.
So when choosing, look at subnet spread and traffic terms first, speed second.
When latency does matter
When the collector sits on a different continent from the platform and there are almost no pauses between requests. Then moving the collector is cheaper than changing addresses.
requests 1,000,000 latency 40 ms ~11 hours of pure waiting latency 200 ms ~55 hours pause 2 seconds ~555 hours, and it decides everything