Recall Forensics

Recall Forensics: The Darwin's Recall That Never Happened — the validation failure behind it

Team ValidLeaf · October 2026 · 6 min read

Recall Forensics is our running teardown of real pet food enforcement actions: we pull the public regulator records and trace each case down to the exact quality-system weakness underneath it. Nothing in this series is invented — every fact below comes from the sources linked at the bottom. This case breaks the pattern of the others we have covered, because there was no recall to examine forensically. Six lots of raw pet food tested positive for dangerous pathogens, the FDA asked the manufacturer to pull them — and the recall never came.

What happened

The U.S. Food and Drug Administration issued an advisory cautioning pet owners not to feed certain lots of Darwin's Natural Pet Products raw cat and dog food, manufactured by Arrow Reliance, Inc. FDA laboratory samples told the story: five samples of the company's raw pet food tested positive for Salmonella, and a sixth tested positive for both Salmonella and Listeria monocytogenes (iHeartDogs, reprinting the FDA advisory).

The implicated products are Darwin's Natural Selections Antibiotic & Grain-Free Chicken Recipe for Cats (lot numbers 10832, 10856, 10890) and Darwin's Natural Selections Antibiotic & Grain-Free Chicken Recipe for Dogs (lot numbers 10828, 10844, 10887). The food ships in frozen 2-pound packages with four separate units, and the brand sells mostly online through a subscription service — meaning the affected product went directly into home freezers across the country (iHeartDogs).

Here is the part that makes this case unusual. The FDA recommended that Arrow Reliance recall all six lots. The firm has not recalled the affected products, and the agency noted the company could continue to distribute them. So FDA did what it does when the manufacturer's lever stays put: it went straight to consumers, telling them to check their freezers for the lot numbers, throw any implicated product away in a secure container, and disinfect every bowl, surface, and storage container the food touched (iHeartDogs).

An honesty note on the timeline: the cited sources do not carry a date stamp for the FDA advisory itself — trade-press reporting shows it was current in mid-September and early October 2026. Whether Arrow Reliance has since initiated a recall is not publicly reported. What the public record establishes is that, as of the advisory, the six pathogen-positive lots had not been recalled by the firm — the FDA advisory itself, rather than a company recall notice, was the basis for the consumer warning (FDA advisory via iHeartDogs).

One fact from the advisory deserves underlining, because it undoes a common consumer assumption: refrigeration or freezing does not kill Salmonella or L. mono. The "frozen" on the package is storage, not a kill step.

The validation gap: a recall plan nobody executed

Most recall stories are about a fix that arrived too late. This one is about a fix that never started. Under the animal food preventive controls rule (21 CFR Part 507), an animal food facility's food safety plan must include a written recall plan — the document that says who decides, who notifies, who traces, and how product comes back. But a written plan is a document, not a capability. When the regulator recommends a recall and the company declines, the hard question isn't whether a binder exists somewhere. It's whether the company has ever run one.

Contrast this with the enforcement case we traced in Raaw Energy's warning letter, where the company recalled repeatedly but never eliminated the contamination's source. RAAW's failure was a recall without a root-cause fix; this is a pathogen finding without a recall at all — the same corrective-action discipline failing at the very first rung. And contrast it with the Walmart Bettergoods frozen pasta case, where a recall did pull the product. The machinery of a recall is supposed to be the one part of the system that works on demand. Here it never turned on.

The subscription model makes the non-recall even more instructive. A direct-to-consumer subscription business holds its customers' contact information and delivery records — arguably the easiest recall-notification channel a manufacturer can have. No retailer chains to negotiate with, no distributor lists to chase. When a firm in that position still does not execute, the gap is not logistics. It is a system-level decision about corrective action — and a readiness assessment would treat it exactly that way: as evidence that the company's CAPA and regulatory-response procedures are untested or unused.

There is a second, quieter gap worth naming: finished-product testing that doesn't act like a control. Six lots were positive across FDA's sampling — five with Salmonella, one with Salmonella plus Listeria. For raw animal food, where no downstream cooking step exists to rescue a contaminated batch, laboratory results on finished product are the last line of defense. Whether the company's own testing ever saw these positives is not publicly known — the record doesn't say. But either the firm's verification testing missed them, or the results didn't drive action. A hold-and-release discipline, or any equivalent verification activity proportionate to the hazard, is what turns a lab result into a decision.

How a GMP Readiness Assessment catches it

This is exactly the kind of failure a GMP Readiness Assessment is built to find before a regulator does — in pet food operations governed by 21 CFR Part 507, current good manufacturing practice and hazard analysis for animal food. Over four weeks, it stress-tests:

• Recall-readiness drills — a mock recall traced from a finished lot all the way to the affected customers, timed. The drill is run against the company's actual distribution model — including subscription shipments and e-commerce order records — not a generic template.
• Tracing architecture — lot codes on every package unit (Darwin's packages carried them, which is precisely why consumers could be told which lots to discard), linked backward to ingredient and process records and forward to shipments.
• Corrective-action and regulatory-response discipline — when a regulator recommends action, is there a documented decision process, a named decision-maker, and a clock? An untested response procedure is where recalls go to stall.
• Verification activities proportionate to the hazard — for raw animal food with no downstream kill step, are finished-product testing, hold decisions, and trend reviews actually wired into shipping decisions?

It closes with a gap assessment, a remediation roadmap, and starter templates — so the recall machinery gets exercised in a drill, not in an advisory.