ScriptManager caches the rejected promise when a resolver throws, so loadScript/prefetchScript never retry #1506

Describe the bug

If a script resolver throws, ScriptManager keeps the rejected promise in scriptsPromises. Every later loadScript / prefetchScript call for the same script returns that same error and never calls the resolvers again. The only way out is to restart the app.

In loadScript and prefetchScript, resolveScript is awaited before the try / finally that removes the entry from scriptsPromises:

const loadProcess = async () => {
  let script = await this.resolveScript(/* ... */); // a throw here skips the finally below

  try {
    // ...
  } catch (error) {
    // ...
  } finally {
    // should delete script promise even script failed
    delete this.scriptsPromises[uniqueId];
  }
};

this.scriptsPromises[uniqueId] = loadProcess();

So the "delete even if it failed" path only runs for loading errors, not for resolving errors. The same pattern is in prefetchScript. It also affects loadScript after a failed prefetchScript, because loadScript awaits the cached prefetch promise and rethrows its error.

Expected: after a resolver error (e.g. the device was offline while the resolver fetched the script URL), the next loadScript call runs the resolvers again, the same way it already retries after a failed native load.

Actual: the next loadScript calls reject right away with the first error, even after the network is back. In our app a user who opened a mini app while offline could not open it again until they killed and reopened the app. A retry button on an error boundary or React.lazy remount did not help.

System Info

React Native 0.79.7 (New Architecture, Hermes)
@callstack/repack 5.2.5, bundler webpack 5, Module Federation v1
iOS and Android, release builds

Re.Pack Version

5.2.5. The code is unchanged in 5.4.2 and on main (32c32c0).

Reproduction

https://github.com/linhh-phv/repack/blob/fix/script-manager-retry-after-resolve-failure/packages/repack/src/modules/ScriptManager/__tests__/ScriptManager.test.ts. The two tests should retry loadScript after resolver failure and should retry prefetchScript after resolver failure fail on main and pass with the fix.

Steps to reproduce

1. Add a resolver that rejects once, then resolves:

   const resolver = jest
     .fn()
     .mockRejectedValueOnce(new Error('Network request failed'))
     .mockImplementation(async (scriptId) => ({ url: Script.getRemoteURL(scriptId) }));
   ScriptManager.shared.addResolver(resolver);

2. await ScriptManager.shared.loadScript('miniApp') rejects with Network request failed, as expected.

3. Call await ScriptManager.shared.loadScript('miniApp') again: it rejects with the same error and resolver was called only once.

In an app: make the resolver depend on the network (for example, fetch the version or URL of a remote), open a federated remote while offline, turn the network back on, then try to open it again.

I have a fix ready and will open a PR: delete the cached promise when resolveScript throws, then rethrow.