been using wretch for some time, nice little lib indeed.

it might be worth tweaking the retry example or defaults to ignore certain http responses that should almost never retry.

we had a case with Zoom API recently where initial POST resulted in 400 Bad Request which retried a few times and subsequently causing 429 Too Many Requests. things like 400 bad request is almost always client's fault!

maybe adding a simple check in the doc or adding a new option / default?

...
retry({
    ...
    // didn't try this yet:
    until: (response, error) => response && (response.ok || ![408, 413, 500, 502, 503, 504].includes(response.status)),
    // or a new option?
    // retryOnAnyResponse: false,
    retryOnNetworkError: false,
    resolveWithLatestResponse: false
  })
...