Handle exceptions gracefully

Make exception handling consistent and safe across the codebase. Audit hotpaths and library boundaries, add try-catch only where the code can recover or must translate exceptions into error codes, ensure destructors are noexcept, and add unit tests for thrown exceptions. Update the coding guide with rules for when to throw, when to catch, and how to propagate errors across threads and through the C API.

Before I roll this out, surface the parts that matter to reviewers: list the modules where exceptions were added or removed, show example call stacks where an uncaught exception could crash the app, and flag places where third-party code throws nonstandard exception types. Suggest minimal test cases reviewers should run to prove behavior is unchanged.

We just lost a game server because an exception on a worker thread escaped and terminated the process. The codebase mixes exceptions and error codes and I cannot find where the thread's work did not catch errors. What are the most likely points of failure to inspect now, what quick code change will prevent process-wide termination while I trace the root cause, and how should I communicate this to ops so they can restart safely?

Across releases we fix crashes by sprinkling try-catch blocks, which slows reviews and leaves inconsistent semantics. Recommend one habit and one repo-level rule that will standardize exception usage, reduce regressions from uncaught exceptions, and make post-crash behavior predictable for players and ops.