Mocha --parallel: test files collide on shared ports and database state when run across workers

Mocha's --parallel mode distributes test files across a pool of worker processes, each loading its files in isolation. That isolation ends at the process boundary: any external resource a suite touches — a TCP port, a database schema, a temp directory — is still shared by every worker on the machine.

The design goal here is to keep an existing suite green under --parallel without rewriting tests, where several files currently assume exclusive access to a fixed port and a single seeded database. Root-level hooks run once per worker process, so setup that was safe in serial mode now executes concurrently, and Mocha itself provides no built-in coordination primitive for namespacing resources between workers.

The uncertainty is which isolation strategy fits Mocha's scheduling model: deriving per-worker namespaces from something stable like the worker's process ID or an environment variable Mocha exposes, versus partitioning resources statically per test file, versus simply accepting serial mode for the coupled subset. Buffered per-file output also makes contention failures harder to attribute.

Assuming a recent Mocha major version (to be confirmed with npx mocha --version):

• Does Mocha expose a documented worker identifier or index that root hooks can use to derive per-worker ports and database names?
• Is there a supported way to pin specific files to serial execution while the rest run in parallel?
• Are there documented limitations of --parallel (reporters, exclusive tests) that should rule it out for suites with shared fixtures?