Every audit came back clean. The strain had been living there a year.
Harvest Table Kitchens did what a careful plant is supposed to do — a current certificate, a swab schedule, a binder of negative results. The one thing the program was meant to catch was the one thing it never looked for. And it was written, plainly, in the plant’s own data.
A plant like this follows a pattern we know well. Not this company — its name, its people, and every number here are invented — but the pattern: proud of its certificate, swabbing every month, certain it was clean.
The organism was in the building the whole time, recorded quietly in the company’s own results — and no one had read those results in the one way that would have shown it.
So we built the case, because the lesson is worth more than the recall it usually arrives with.
- Your food-safety system runs through one very capable person.
- Your environmental program is monthly, weighted to floors and drains, mostly clean.
- A swab site has come back positive more than once — cleaned and re-swabbed until it cleared.
- You hold a current certificate, and treat it as evidence the plant is in control.
If two or more are true, this case is describing your operation, not a stranger’s.
The file, as it arrived
Nine records. Every one internally consistent. Every one, read on its own, defensible.
A plant like this fails between documents, never inside one. Open the file and read what you’d catch — then see what we found, just below.
What the records showed, together
No single page is alarming — read apart, each record is defensible. Read together, by site and across time, the same evidence resolves into four threads, all converging on a single lot.
A program built to pass an audit, not to find harborage.
Monthly swabbing, weighted to floors and drains. Food-contact surfaces only quarterly. Every isolate logged as Listeria spp. — present or absent (EMP-001). A positive meant a re-clean and one negative re-swab, never swabbing outward to find where it lived.
A resident strain, written in the plant’s own results.
A drain and a condensate point tested positive month after month for fourteen months. Each was closed “sanitized → ND” (EMP-LOG-26). Never speciated, so never named. Persistence by site was the one signal that mattered — and the one the program never used.
A control that was never really a control.
Cooked and raw components were assembled cold in that same wet room — after the only lethal step. The plan called the Listeria hazard “controlled by sanitation” (FSP-001). That is an unvalidated routine standing in for the validated control a RTE food requires.
A “fresh” date that gave it time to grow.
The 14-day shelf life came from a sensory panel, not a growth study. But Listeria grows cold — so the same “fresh” window that sells the product gives a low initial load time to reach an infectious dose. The sensory hold conceals the growth, because spoilage organisms never track Lm.
Read apart, every record was defensible. Read together — by site, across time — they name a resident strain, an unvalidated control, and a shelf life that let it grow. The production record, the open EMP alert, and the below-target wash-water log all land on the same unit: lot L 26059 A. That convergence is the finding — a 21 CFR 117 ready-to-eat operation whose Listeria program was built to clear an audit, not to read its own data.
None of it needed a new test to see. It needed someone to read the records the plant already had — the way they were meant to be read.
What was at stake
A recall and a Form FDA 483
The lot shipped on area data with an open alert. A consumer was hospitalized; the inspection WGS-matched a single resident strain across dates and wrote it up — after the certificate on the wall said the plant was in control.
The program reads its own data
Read by site across time, the persistence was visible months earlier — nameable, contained, and fixed before a lot ever left on incomplete evidence. The same records, read the right way, are the early-warning system.
If this is your operation
You don’t need us to find out whether your program has the same blind spot. Most of it you can check yourself, this week, with records you already have.
- Sort your last 12–18 months of results by site, not by month.Does any single location keep reappearing? A “percent-positive” that looks in control can hide one spot that never left.
- When a swab is positive, what happens next?Do you clean and re-swab until it clears — or swab outward to find where the organism is living?
- Is anything you find speciated and subtyped?You can’t recognize a resident strain you never named. Genus-level present/absent is structurally blind to persistence.
- Where did your “fresh” date come from?A challenge and growth study — or a tasting panel? For a refrigerated RTE, only one of those is a safety number.
Nothing here is built around this one plant — each door points somewhere bigger, and no one of them is the “right” one. The case is just where the three meet.
A clean audit was never the same thing as a clean plant.