fix(share-tokens): single-flight load and serialized atomic persist (Closes #102)

[SakethSumanBathini]

Closes #102

Problem

backend/src/services/shareTokenStore.ts has two independent races, and the first is considerably more damaging than it looks.

1. load() can read more than once. The guard is if (cache) return cache, but cache is only assigned after the await fs.readFile(...) resolves. Every caller arriving during that window sees null and starts its own read, so each ends up holding a different object. Each then adds its token to its own copy and persists it — and the last write wins, discarding every other caller's token.

2. persist() writes concurrently to one path. fs.writeFile truncates and then writes, so two overlapping calls can leave a partially written file. That matters more than a normal torn write would, because load() handles a parse failure with a bare catch { cache = {} } — a corrupted file is not surfaced anywhere, it silently becomes an empty token store.

Reproduction

I transpiled the current file and drove 300 concurrent createShareToken calls against it:

cold cache | issued 300 | persisted 1/300 tokens   *** 299 LOST ***

All 300 callers hit the cold-cache window, so all 300 wrote single-token files over each other.

The same run against the patched file:

cold cache | issued 300 | persisted 300/300 tokens   OK

To be precise about scope: with a warm cache the current code is fine (301/301), because all callers then share one object. The severe path is specifically the cold-start burst — process start, or the first requests after a restart.

I should also be straight that I could not reproduce a torn write on my filesystem; ~64KB writes completed without visible interleaving. The atomic rename below removes that failure mode by construction rather than by observation.

Change

Single-flight load. The in-flight promise is memoised, not just its result, so concurrent callers await the same read and receive the same object:

let loadPromise: Promise<TokenCache> | null = null;

const load = async (): Promise<TokenCache> => {
  if (cache) return cache;
  if (!loadPromise) {
    loadPromise = (async () => { /* read, parse, assign cache */ })();
  }
  return loadPromise;
};

Parsing is also shape-checked now — a value that isn't a plain object is treated as empty rather than returned as a TokenCache.

Atomic write. writeAtomic writes to a randomly-named temporary file in the target directory (so rename stays within one filesystem), flushes the handle, then renames over the target. rename is atomic, so a reader sees either the whole old file or the whole new one — never a partial. The random suffix stops two writers colliding on the temp path, and a failed write removes its own temp file.

Serialized persists. Writes chain onto a queue so only one runs at a time. The queue advances with a swallowed rejection so a single failed write doesn't wedge every later one, and persist still logs rather than throws — the original contract is unchanged.

Scope

This addresses in-process concurrency, which is what the issue describes. It does not add a cross-process lock: two backend processes sharing PROOFDESK_DATA_DIR could still lose an update, though the atomic rename means they can no longer corrupt the file. docker-compose.prod.yml runs a single backend, so that isn't currently reachable — happy to add file locking if you'd like it covered.

Verification

• npx tsc --noEmit in backend/ clean.
• 300-way concurrency test above: 1/300 before, 300/300 after.
• Confirmed no .tmp files are left behind after the run.

Note: npm test is red on clean main independently of this branch.

[coderabbitai]

Warning

Review limit reached

@SakethSumanBathini, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 3 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 458102dd-b0bf-48c7-a5f4-86fbf0e88178

📥 Commits

Reviewing files that changed from the base of the PR and between 2f06e6e and 048f092.

📒 Files selected for processing (1)

• backend/src/services/shareTokenStore.ts

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}