Our server currently mixes try/catch and ad-hoc error handling across request, auth, and accounting layers. This makes control flow unpredictable and complicates typed error propagation.

We should standardize on neverthrow for all expected failure paths (validation, auth, quota, accounting, upstream 4xx/5xx, etc.), and reserve throw only for truly exceptional cases (basically never).

Bounty details

Total prize: $3,000

This is a fairly large change, AI slopping this up will certainly not work. This will require deep diving the existing server structure. If you want to work on this you should demonstrate your knowledge of production grade TS before starting on it, you will gain approval from the team to take a stab at it.

You will have access to me and the team for questions + design decisions. There will be behavioral changes to get team consensus on here as the current implementation has many known bugs.