test: fix flaky timing and correctness issues in reporter/plugin manager tests

[Copilot]

Four correctness and reliability issues in the test suite identified during review of #130.

observability/reporter_test.go

• TestReporter_StopFlushesRemaining: handler was setting received=true for any request, meaning the async registration POST to /api/v1/admin/instances/register could satisfy the assertion before the actual flush to /api/v1/admin/ingest/executions ever happened. Now only the ingest path triggers the flag.

• TestReporter_BufferAndFlush: replaced fixed time.Sleep waits with a deadline/poll loop — exits as soon as all three ingest paths are observed, fails fast with a descriptive message after 2s. Added t.Cleanup(cancel) to guarantee goroutine teardown.

deadline := time.Now().Add(2 * time.Second)
for {
    mu.Lock()
    execFlushed := receivedPaths["/api/v1/admin/ingest/executions"] > 0
    // ...
    mu.Unlock()
    if execFlushed && logsFlushed && eventsFlushed { break }
    if time.Now().After(deadline) {
        t.Fatalf("timed out waiting for reports to be flushed, got: %#v", receivedPaths)
    }
    time.Sleep(10 * time.Millisecond)
}

plugin/manager_test.go

• DB setup: four tests were opening SQLite directly via sql.Open("sqlite", ":memory:") + defer db.Close(), bypassing the existing openTestDB(t) helper in native_test.go. Consolidated to use the helper (which registers cleanup via t.Cleanup). Removed now-unused database/sql and _ "modernc.org/sqlite" imports.

• TestPluginManager_AllPlugins: Register/Enable calls were silently discarding errors with _ = .... Replaced with t.Fatalf so a future regression points directly at the failing call rather than a downstream assertion.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.