Slow writes on Upstash Global: every write goes to the primary region first

# Diagnosing slow writes on an Upstash Global database

## Symptom

Reads are fast worldwide but writes are slow, especially from regions
far from the primary.

## Cause

Every write travels to the primary region, gets processed, then
replicates out to read regions. There is no local-write path. The
docs state the tradeoff openly: Global optimizes read latency and
may not suit write-heavy use cases.

## Confirm

1. Check which region is primary in the console.
2. Measure write latency from your app region versus from the
   primary region. If primary-region writes are fast and yours are
   slow, it is distance, not load.
3. Check your write/read ratio. Above roughly 30% writes, you are
   in write-heavy territory for this architecture.

## Fix options

- Move the primary region to where most writes originate.
- Split the workload: a regional database for the write-heavy
  parts, Global for the read-heavy parts.
- If writes must be fast globally and consistent, Upstash Global is
  the wrong tool. Say so early rather than tuning around it.

## Verify

Write P99 from the app region drops to near the published same-
region baselines after the primary move or the split.