fix(ci): cap back-end jest workers to prevent OOM on CI runner

[diegomayorga-dept]

Title*

fix(ci): cap back-end jest workers to prevent OOM on CI runner

Type of Change*

• Bug fix

Description

The Build & Deploy Back-End job in cicd-release-quality (and the sandbox/UAT release pipelines) started failing deterministically on development after PR #823 (fix(EGV-206): upgrade aws-cdk-lib), which bumped aws-cdk-lib from ^2.176.0 to ^2.260.0.

Root cause is an out-of-memory kill, not a test failure:

• The packages/back-end/test/infra/** suites synthesize full CDK stacks (e.g. easy-genomics-nested-stack, auth-nested-stack), each consuming ~1.4 GB of heap.
• Back-end Jest runs with default parallelism (cpuCount - 1 → 3 workers on the 4-vCPU ubuntu-latest runner), each worker inheriting NODE_OPTIONS=--max-old-space-size=8192 and running with v8 coverage enabled.
• The larger aws-cdk-lib raised each synth's footprint just enough that the aggregate RSS across the parallel workers crossed the 16 GB runner limit. The OS OOM-killer then SIGKILLed the process group — which is why the run exits 1 with no Jest summary, output cut off mid-flush, and no V8 "Reached heap limit" message.

This is why the failure was deterministic and re-running the pipeline did not help.

The fix reduces peak memory without touching application code, all via the projen source of truth (.projenrc.ts) which regenerates the back-end config:

• maxWorkers: '50%' — caps concurrent Jest workers so the heavy infra synth suites don't all run at once (2 workers on the CI runner instead of 3).
• --workerIdleMemoryLimit=2GB — recycles a worker once it exceeds 2 GB, freeing accumulated coverage/synth memory between suites instead of letting it grow for the worker's lifetime.

Testing*

• Ran the full back-end suite locally with coverage enabled and the new flags. Before the fix the run was OOM-killed (exit 1, no summary); after the fix it completes cleanly: 66/66 test suites passed, 491/491 tests passed, exit 0, with per-worker heap bounded at ~1.4 GB (workers recycle at the 2 GB idle limit rather than accumulating).
• Verified projen regenerated the config correctly:
  • packages/back-end/package.json → jest.maxWorkers: "50%"
  • packages/back-end/.projen/tasks.json → jest --workerIdleMemoryLimit=2GB --passWithNoTests --updateSnapshot
• No test code changed; all existing tests still pass.

Impact

• CI/build only — unblocks the back-end build/deploy pipeline. No runtime, infrastructure, or application behavior changes.
• Back-end test wall-clock may increase slightly on CI due to reduced parallelism, in exchange for a stable, non-OOM run.
• No new dependencies.
• Follow-up (out of scope): the underlying driver is the ever-growing infra synth footprint (same trend behind the nested-stack resource-limit work). Splitting the nested stack remains the durable path; this change buys headroom now.

Additional Information

• Changed files: .projenrc.ts (source), plus the two projen-generated files packages/back-end/package.json and packages/back-end/.projen/tasks.json. Regenerate with pnpm exec projen if editing .projenrc.ts.
• --workerIdleMemoryLimit is passed via extraCliOptions because it is not exposed in projen's JestConfigOptions type; maxWorkers is set directly in jestConfig.

Checklist*

• No new errors or warnings have been introduced.
• All tests pass successfully and new tests added as necessary.
• Documentation has been updated accordingly. (N/A — CI config change)
• Code adheres to the coding and style guidelines of the project.
• Code has been commented in particularly hard-to-understand areas.