The new diagnostic interface has revealed some problems on the metadata wrangler that stop the circulation manager's NYT best-seller update script from running to completion.

That script relies on the metadata wrangler to convert an author name string as found in the NYT API (e.g. "Bill O'Reilly and Martin Dugard") into a canonicalized author name (e.g. "O'Reilly, Bill") that can be used to identify a book in the library's collection via a title/author lookup. It's a creaky system but it works most of the time.

However, sometimes something goes wrong on the metadata wrangler's side, and when it does, the entire NYT update script crashes.

There are three things we could be doing better:

1. This URL gives a 500 error because the "ISBN" is actually an Amazon ASIN. Someone at the NYT put the wrong info into their spreadsheet, or maybe the book has an ASIN but no ISBN. The metadata wrangler should catch this error and proceed as if no ISBN had been specified.
2. Sometimes the canonical-author-name request just times out and the metadata wrangler gives a 500. This might be related to load on the metadata wrangler, in which case we need to boost the hardware, or it might be a performance problem that happens in rare cases, in which case we need to investigate and fix the problem.
3. If the metadata wrangler gives a 500 on a canonical-author-name request, the circulation manager's script should not crash. It's more important that the script run to completion than that we get a canonical name for every single author.