Skip to content

Case Study: HL7 AU Sparked Testing Event

A real connected-testing event doubled as a stress test for FHIR Frog itself — and surfaced three genuine library bugs that unit tests alone hadn't caught.

The Suite

Ahead of the HL7 AU Sparked Testing Event (8 July 2026), a 20-test FHIR Frog suite was built to prove four scenarios end to end:

  • Scenario 1 — eRequest/eReferral, including provider lookup
  • Scenario 3 — Patient Summary exchange
  • Scenario 4 — medication dispense reconciliation
  • Scenario 5, Step 1 — provider search (reused directly from Scenario 1)

Tests ran against a mix of targets: ADHA's public HCPD sandbox (read-only) and local HAPI FHIR servers — the same "shared infrastructure for read paths, local servers for anything that mutates state" split FHIR Frog's own sample suites use.

What It Found

Running real TestScripts against a real, external, spec-compliant sandbox — rather than a permissive local test server — surfaced gaps that a more forgiving target had let slide:

  • search operations were ignoring params entirely. Every bare, unparameterized search FHIR Frog issued was rejected by the sandbox, which (correctly, per spec) refuses unparameterized searches. See Search Examples for how params is now forwarded.
  • update always reported 200, even when the server returned 201. An eRequest lifecycle test asserting "created" failed despite the update genuinely succeeding. See the note on Operations → Update.
  • There was no way to attach custom headers or auth to outgoing requests at all. Needed both for a sandbox-specific header requirement and as a building block for the SMART App Launch flow that Scenario 2 will need next. See Authentication for the resulting requestInterceptor hook and fhir-frog-smart module.

All three were genuine correctness bugs in OperationExecutor, not test-suite mistakes — fixed in the library, with new regression tests, rather than worked around per-consumer.

Takeaway

Dogfooding FHIR Frog against a real external sandbox under real event deadlines found gaps that permissive local test servers and unit tests hadn't. None of the three fixes were scenario-specific — every one benefits any TestScript author, not just this event's suite.

Next Steps