Summary

When performing sweeping changes to a monorepo, for example upgrading TypeScript, one will often find themselves in the following loop:

1. Run a broad rush build or rush test command to find issues
2. Encounter several projects with issues
3. Drill into a single project's failure log and work through the issues
4. Recompile the failed project and see if any new issues show up
5. Once the focused project is fixed, go back to (1) because all record of which other projects failed is hard to come by

It would be useful to be able to run a command like rush test --show-existing-failure-logs that would simply build the graph, look for the error log files, and for any operations where they exist, replay the log in the same manner as for a build cache hit. All operations that did not have existing failures would be silenced.

Additional use cases:
AI agents could run rush build --to x, ignore the output (other than exit code), then run rush build --to x --show-existing-failure-logs to inject only the relevant information into the context.