UML `alt` Fragment vs Per-Message Guards for Deterministic Sequence Traces

Goal: generate deterministic executable traces from a UML 2.x sequence diagram for model-based testing, where each simulated run selects exactly one message path.

Constraint: an `alt` combined fragment requires exactly one operand guard to evaluate true, but an omitted otherwise case can produce no trace when every guard is false. Per-message guards avoid grouping overhead, yet UML does not standardize their expression language or evaluation order, so multiple true guards or missing tool resolution can create ambiguous traces.

Uncertainty: whether to model alternatives as an `alt` fragment with interaction constraints or as independent per-message constraints, since tool support for executing raw per-message guards in code-generation pipelines is less consistent than for `alt` simulation.

Questions:

• Does an `alt` fragment provide stronger trace determinism than per-message guards when guards are expressed in OCL versus a tool-specific action language?
• Should an `otherwise` operand be required in an `alt` fragment to guarantee at least one executable trace under model-based testing?
• Which approach better preserves guard semantics when exporting the sequence diagram to a test execution engine?