The question I want this model to answer is not “exactly how long will the hardware take?” Relay cannot know that statically. The useful question is “if this visible wait is reachable, which declared timing expectations could it disrupt?”
Relay evaluates literals, signed values, common C numeric suffixes, simple named constants,
arithmetic, and min/max. Language profiles also understand common duration constructors such
as Go time.Second and Rust Duration::from_millis.
A duration is compared only with a task whose configured function can reach that call. The report
can show how many declared periods overlap the wait and how far a declared latency budget is
exceeded. execution_context is printed as provenance; Relay does not infer schedulers, threads,
processes, event loops, or cores.
These are static review calculations, not response-time analysis or measured worst-case execution
times. Release times, execution cost, priorities, pre-emption, interrupts, multicore scheduling,
blocking overlap, libraries, I/O variability, and hardware are not modelled. That is why the
command is named summary, not timeline or simulate.