Edge Functions: fresh workers on sequential requests

Hi, I’m trying to understand worker/isolate reuse behavior in Supabase Edge Functions.

Setup:

• Supabase Edge Functions
• Hono
• direct Postgres access from the function (via Drizzle / SUPABASE_DB_URL)
• tested with strictly sequential load only (concurrency=1)

What I see:

• locally (supabase functions serve), workers are reused as expected
• in production, many sequential requests still seem to hit a fresh worker/isolate
• production logs often look like:

booted (time: 130-170ms) [worker id:xxxxxx] started [worker id:xxxxxx] accepted request: 1 [worker id:xxxxxx] request #1 processed for 100-700ms

So for many requests I see:

• a fresh boot
• a fresh worker id
• accepted request: 1

Locally, reuse looks normal:

[worker id:abc123] accepted request: 1 [worker id:abc123] accepted request: 2 [worker id:abc123] accepted request: 3

Questions:

1. Is it expected that production Edge Functions frequently create a fresh worker even with concurrency=1?
2. Is worker/isolate reuse always best-effort only, with no way to improve affinity/reuse?
3. If a function uses direct Postgres access, should I assume in-memory caches / pooled connections inside the worker are not reliable optimizations in production?
4. For latency-sensitive endpoints, is the recommended approach to avoid relying on worker reuse entirely?

Thanks! I can share more logs if helpful.