fix(server): skip undecodable provider runtime rows when listing sessions

[t3dotgg]

Problem

A single stale row in provider_session_runtime — e.g. one written by an older build with a status value that's no longer in ProviderSessionRuntimeStatus — made ProviderSessionRuntimeRepository.list() fail wholesale. That silently broke every consumer that enumerates sessions, most notably ProviderSessionReaper, whose sweep then failed on every run:

WARN provider.session.reaper.sweep-failed
  cause: PersistenceDecodeError ProviderSessionRuntimeRepository.list:decodeRows

Seen in the wild on a machine with one leftover status = "ready" row from June: idle-session reaping was completely disabled until the row was deleted by hand.

Fix

list() now decodes rows individually: a row that fails to decode is logged (provider.session.runtime.row-skipped, with the threadId and the decode error) and skipped, instead of failing the whole list. Point reads (getByThreadId) intentionally keep surfacing decode errors, so a corrupt row for a specific thread is still visible where it matters.

Updated the correlation test that pinned the old fail-the-whole-list behavior to assert the bad row is skipped and valid rows are still returned.

Verification

• pnpm typecheck in apps/server
• pnpm test src/persistence/RepositoryErrorCorrelation.test.ts src/provider/Layers/ProviderSessionDirectory.test.ts src/provider/Layers/ProviderSessionReaper.test.ts — all passing

🤖 Generated with Claude Code

---

Note

Medium Risk
Changes enumeration semantics for session consumers (e.g. reaper) so stale rows are silently omitted rather than blocking the sweep, while targeted reads still fail on corruption.

Overview
ProviderSessionRuntimeRepository.list() no longer fails the entire enumeration when one DB row fails schema decode (e.g. legacy status values). Each row is decoded in isolation: failures emit provider.session.runtime.row-skipped with threadId and a safe error message, and valid rows are still returned.

getByThreadId is unchanged—decode errors for a specific thread still surface as PersistenceDecodeError.

The persistence correlation test now asserts that a corrupt row is omitted from list() while a valid upserted row remains visible; SQL failure correlation behavior is unchanged.

^{Reviewed by Cursor Bugbot for commit 3f61538. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Skip undecodable rows in ProviderSessionRuntimeRepository.list instead of failing

• ProviderSessionRuntimeRepository.Service.list now maps each row to Option.some on success or Option.none on failure, then filters out failed rows rather than propagating a PersistenceDecodeError for the entire operation.
• Failed rows emit a warning log with the threadId and a safe error message.
• The test in RepositoryErrorCorrelation.test.ts is updated to assert that only valid rows are returned when bad rows exist.
• Behavioral Change: runtimes.list() now returns a partial result instead of failing when any row is undecodable.

Changes since #3951 opened

• Modified make.list function in ProviderSessionRuntime to skip rows that fail decoding [3f61538]
• Added module imports for functional error handling utilities [3f61538]

^{Macroscope summarized 2f3904a.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 4d9eb82e-97ba-4ea5-8cbb-cd9104359a9e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/skip-undecodable-provider-runtime-rows

---

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

[macroscopeapp]

Approvability

Verdict: Approved

Straightforward defensive bug fix that changes decode error handling from failing the entire list operation to logging a warning and skipping undecodable rows. Clear intent, limited scope to one repository method, with updated test coverage.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Effect Service Conventions: one finding on log payload safety in ProviderSessionRuntime.list. See inline comment.

^{Posted via Macroscope — Effect Service Conventions}

Dismissing prior approval to re-evaluate 2f3904a

Dismissing prior approval to re-evaluate 3f61538