[TEST] Add chaos tests for task execution failures and partial artifacts

[utksh1]

Problem

SecuScan needs a production-grade improvement in this area: Failure-mode coverage..

Scope

Inject process crashes, parser exceptions, DB write failures, and partial report files to verify cleanup and task status behavior.

Acceptance Criteria

• The implementation is focused and does not introduce unrelated UI, docs, lockfile, or formatting churn.
• Security-sensitive behavior has explicit negative tests where applicable.
• Existing tests continue to pass, and new tests cover the main success and failure paths.
• Documentation or configuration examples are updated when operator behavior changes.

Verification

Tests should assert no stuck tasks, orphan artifacts, or misleading success states.

Difficulty

Hard, useful issue intended for experienced contributors.

[VIDYANKSHINI]

I would like to work on this issue.

I am interested in contributing to the backend testing and failure-handling improvements described in this task. I can work on adding focused chaos tests covering process crashes, parser exceptions, database write failures, and partial artifact/report generation scenarios to verify correct cleanup behavior and task status handling.

Planned scope:

• Add failure-mode and recovery test coverage
• Verify no stuck tasks or orphan artifacts remain after failures
• Test negative/error scenarios for security-sensitive behavior where applicable
• Ensure misleading success states are not produced
• Keep the implementation scoped without unrelated formatting or UI changes
• Run and validate existing as well as newly added test cases

Please assign this issue to me if it is still available.

[divyansha12]

Hi @utksh1, GSSoC '26 contributor here and would love to work on this!

One thing I want to get right before starting: chaos tests are only useful if they're deterministic. My approach would be to freeze failure injection points using monkeypatch so tests never rely on actual timing or external state. For DB write failures I'd patch at the ORM/query layer not the connection level so the failure is clean and repeatable. Partial artifact tests would write incomplete files to tmp_path and assert the task ends in a failed/cleaned state rather than stuck or falsely succeeded.

Can you assign this to me?

[HarshParmar029]

Hi @utksh1,

I noticed the issue is already assigned and an initial PR has been linked. I’d still be happy to contribute if there’s any remaining scope around additional chaos scenarios, edge-case coverage, cleanup verification, or test hardening.

I can help with:

• deterministic failure injection tests
• orphan artifact/task cleanup assertions
• parser/DB failure edge cases
• flaky-test prevention and isolation improvements
• reviewing uncovered execution paths

Please let me know if there’s a specific gap or follow-up task I can take up. Thanks!