We found an unexpected telemetry join issue while comparing device IDs extracted from mongosh data with Atlas CLI device IDs: there was no overlap in the sample that was checked, even though the expectation was that CLI/shell overlap should be high on the same machine.

Expected behaviour:
Every tool running on the same machine reports the exact same device id, byte by byte.

If possible ensure this with tests that would prevent regressions.

Open questions

• to clarify with analytics: should we keep the old device id around during transition?
• based usage of the package from community: should this be a breaking change for the library users (mayor bump), or should this be a format: 'option', keeping the old algorithm as default?