Exponential-backoff retry for transient failures — a 500, 502, 503 or
504, or a network failure (see isTransientFailure). Wraps the attempt
with withRetryBackoff using the caller-provided telemetry
hook for every retry tick.
This policy only makes sense on idempotent verbs (typically GET):
retrying a POST that may have landed server-side is a duplicate
write in disguise. The consuming client gates application of this
policy accordingly; the policy itself is neutral about method.
Ownership: transient 5xx and network failures. Any other 5xx (501,
505), a caller's abort or any other error propagates untouched on the
first failure — no retry.
Exponential-backoff retry for transient failures — a 500, 502, 503 or 504, or a network failure (see isTransientFailure). Wraps the attempt with withRetryBackoff using the caller-provided telemetry hook for every retry tick.
This policy only makes sense on idempotent verbs (typically GET): retrying a POST that may have landed server-side is a duplicate write in disguise. The consuming client gates application of this policy accordingly; the policy itself is neutral about method.
Ownership: transient 5xx and network failures. Any other 5xx (501, 505), a caller's abort or any other error propagates untouched on the first failure — no retry.