Severity: Medium
Observation
DatabricksConnectionTest.CanDetectConnectionParameterErrors runs 93 parameterized sub-cases that intentionally throw ArgumentException / ArgumentOutOfRangeException from the connection-parameter parser (bad timeout values, non-bool strings for bool flags, etc.). All 93 pass — but the archive produced zero spans for those failure paths.
In the same archive, the 1× TrySetGetDirectResults_UsesDatabricksDefaultGetDirectResults test produced ~20 spans. So the test suite IS tracing — but the parameter-validation throws happen synchronously, before any Activity is opened.
Contrast: DriverTests.CanDetectInvalidServer / CanDetectInvalidAuthentication DO produce Status=Error spans on HiveServer2Connection.OpenAsync / SendAsync because those failures happen inside instrumented async methods. So the gap is specifically the synchronous-throw-before-Activity code path.
Why it matters
For one-off debugging, the caller catches the ArgumentException and sees the message — limited value to also trace it. But for fleet-level observability ("X% of connection attempts fail at param validation"), you need a span with Status=Error to aggregate on. Today that signal is invisible.
Suggested fix
Wrap the parameter-parser entry point in an Activity (or just the throw site). Even a thin "validation" span with Status=Error and exception.type = ArgumentException would close the loop for fleet metrics.
Evidence
Traces.DatabricksConnectionTest.20260527_144209/ — 93 invalid-param tests, 0 spans emitted for them; 515 spans total all Status=OkTraces.DriverTests.20260527_153928/ — invalid-server / invalid-auth tests DO emit Status=Error spans (different code path)
Severity: Medium
Observation
DatabricksConnectionTest.CanDetectConnectionParameterErrorsruns 93 parameterized sub-cases that intentionally throwArgumentException/ArgumentOutOfRangeExceptionfrom the connection-parameter parser (bad timeout values, non-bool strings for bool flags, etc.). All 93 pass — but the archive produced zero spans for those failure paths.In the same archive, the 1×
TrySetGetDirectResults_UsesDatabricksDefaultGetDirectResultstest produced ~20 spans. So the test suite IS tracing — but the parameter-validation throws happen synchronously, before anyActivityis opened.Contrast:
DriverTests.CanDetectInvalidServer/CanDetectInvalidAuthenticationDO produceStatus=Errorspans onHiveServer2Connection.OpenAsync/SendAsyncbecause those failures happen inside instrumented async methods. So the gap is specifically the synchronous-throw-before-Activity code path.Why it matters
For one-off debugging, the caller catches the ArgumentException and sees the message — limited value to also trace it. But for fleet-level observability ("X% of connection attempts fail at param validation"), you need a span with
Status=Errorto aggregate on. Today that signal is invisible.Suggested fix
Wrap the parameter-parser entry point in an Activity (or just the throw site). Even a thin "validation" span with
Status=Errorandexception.type = ArgumentExceptionwould close the loop for fleet metrics.Evidence
Traces.DatabricksConnectionTest.20260527_144209/— 93 invalid-param tests, 0 spans emitted for them; 515 spans total allStatus=OkTraces.DriverTests.20260527_153928/— invalid-server / invalid-auth tests DO emit Status=Error spans (different code path)