Thick borg (future): an app per common migration story #856

Future plan. Slim borg checks each primitive once. Thick borg is a decent-sized app per common migration story, run through the same kind of suite on each target. Each app is also the example for its story.

Split the stories by three separate choices (where the code runs, which engine, who hosts the data) and write one adapter per choice, never one per combination.

Stories, most common first:

• Keep the database, move the code: a Node or Next app on Postgres (Neon) moved to a Worker through Hyperdrive, and to a Hetzner box. Needs a postgres binding: Hyperdrive on the edge, a direct connection on a box, PGlite as the desktop twin, and a Neon branch per test stack.
• Leave a PaaS for a box: Railway or Heroku to Hetzner, with sqlite or Postgres on the volume
• Go all-edge: Postgres to D1, with a lint that flags what D1 won't run (BEGIN, ATTACH, too many bound parameters)
• Lambda + DynamoDB to Workers + KV / Durable Objects
• Mongo: probably "keep Atlas, move the code". Check whether FerretDB works as a twin.
• Better Auth on all three targets: a secret minted per install on desktop, and login handled by the edge with the session handed to the desktop
• One sync engine (Yjs in a Durable Object, or Replicache / LiveStore on D1)

Alchemy beta.77 already has Neon, PlanetScale, Hyperdrive, AWS, Fly, Docker and Kubernetes. It has no DigitalOcean, Mongo or Supabase.

Free where possible: Cloudflare and Neon free tiers. PlanetScale and the boxes cost cents per run, so run them rarely.