Started thinking node selection was mostly a distance problem. It turned out to be a network quality problem instead.

During one inference run, the nearest region looked like the obvious choice. The requests reached the node quickly, but completion times became inconsistent. Some jobs finished immediately while others waited long enough to trigger retries.

The surprising part was that a more distant route stayed stable from start to finish. It wasn’t the extra distance that mattered. It was fewer routing changes, lower congestion, and more predictable verificattiming.#OPG

That changed how I think about OpenGradient infrastructure. Geographic proximity is useful, but it shouldn’t outweigh real network conditions. Stable routing, consistent verification latency, and queue health often matter more than being the closest node on the map.

Distance is still part of the decision. It just isn’t the deciding factor anymore.

What should @OpenGradient prioritize most for node selection: network stability, verification latency, queue health, or geographic proximity?

$OPG

$ACT

$MANTA