Analyzed Failure
A failure that has been recognized and classified by the experiment infrastructure.
Carries a stable resolvedId (used to group identical failures across runs), a human-readable description, a FailureKind bucket, and a TransiencyLevel consumed by the retry policy. The recognizers that produce these live in the consuming application — the library only owns the vocabulary.
Constructors
Builds an AnalyzedFailure from an exception, capturing its stack trace into originalStackTrace.
Properties
Human-readable description of what went wrong.
High-level FailureKind bucket this failure belongs to.
Stack trace of the originating exception, captured as a string for record-keeping.
Stable identifier used to group identical failures across runs.
Transiency assignment consumed by the retry policy. Recognizers set this honestly based on what the exception alone can reveal — duration variance ("a transient outage can last seconds or hours") is handled by the retry strategy, not by the enum.