An architecture review that cancelled a rewrite

A three-week review that replaced a planned eighteen-month rewrite with four months of targeted work.

Orbital had board approval for a rewrite and a slide deck explaining why the current system could not scale. The deck was well argued and, as it turned out, wrong about where the time was going.

Three weeks of measurement found that 94% of dispatch latency sat in a single N+1 query path behind a cache that had silently stopped being warmed in 2024.

The review recommended four months of targeted work instead. The rewrite was cancelled. Two of the five proposed services were built; the other three were never needed once the latency was gone.