Local static build exits 0 and builds nothing when the Tina datalayer is unreachable

[jamiefolsom]

Running astro build with STATIC_BUILD=true outside of tinacms dev (and without a reachable MongoDB) produces no pages, no error, and exit code 0. The log ends mid-build, right after a content loader starts:

11:20:47 [content] Syncing content
11:20:47 [i18n-loader] Fetching data.
(process exits 0)

What happens: the loaders query TinaCMS through @tina/databaseClient, which talks to the datalayer directly. With no datalayer available the query never settles, node runs out of pending work, and the process exits cleanly in the middle of the await. In a script or CI job this looks like a successful build unless something inspects dist/.

The build should fail loudly when the datalayer can't be reached — a connection timeout or an upfront reachability check in the loaders would both do it.

Workaround for local static builds: run the build inside the datalayer, TINA_PUBLIC_IS_LOCAL=true tinacms dev -c "astro build". (On Netlify this doesn't bite because MongoDB is reachable at build time.)

Found while running a static build of marronage staging locally (performant-software/dhtools#17).

Part of #710. Fixing this lets us close #752.

[jamiefolsom]

One correction to the workaround: the build has to run in production mode, so the full command is TINA_PUBLIC_IS_LOCAL=true tinacms dev -c "NODE_ENV=production astro build". Without that, tinacms dev runs the build in development mode, and search does not work on the site it produces. #752 describes that failure.

[ajolipa]

I'm a bit confused about what the desired behavior here is; in what situation are we attempting to run astro build without access to a Tina data backend? Is that something we intend to support? Or is the idea just to make sure the error that's thrown if someone tries that for some reason is more informative?