Reduce Vercel DB pool sizes and enforce statement_timeout

[iscekic]

Summary

• Reduce pool max from 100 to 10 per pool (primary + replica) — at peak, dozens of warm Vercel instances across 20+ regions each holding max:100 connections contributes significantly to hitting the Supabase PgBouncer max_client_conn limit (3000).
• Apply POSTGRES_MAX_QUERY_TIME as statement_timeout on the pool — the env var was validated to exist on startup but was never actually passed to the pool config.

Part of the fix for max_client_conn connection exhaustion (see also #497 for the Cloudflare worker changes).

[iscekic]

This PR contains the Vercel Next.js app changes that were originally part of #497. They were split out so the Cloudflare worker changes (replacing pg.Pool with per-operation pg.Client to match the Hyperdrive pattern) could be reviewed and deployed independently from the Vercel pool size reduction.

[kiloconnect]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Notes

• Pool size reduction from 100 → 10 is appropriate for Vercel's serverless model where many warm instances each hold connections.
• Adding statement_timeout correctly wires up the POSTGRES_MAX_QUERY_TIME env var that was already validated at startup (line 18) but never actually passed to the pool config — good catch.
• Both primary and replica pools receive identical changes, maintaining consistency.

Files Reviewed (1 file)

• src/lib/drizzle.ts - 0 issues

[iscekic]

Note to self - statement_timeout doesn't work