Foreign Material Recalls in Food and Drink: A Monthly Tracker
In short: A foreign material recall pulls food that may contain metal, glass, plastic, rubber, stones or another object that should not be in it. This tracker lists the 15 foreign material recalls published by the US FDA, Canada's CFIA and Food Standards Australia New Zealand from August to mid-November 2026, and reads what set the size of each one.
- Metal led the 15 cases (7), then glass (3) and plastic (2). FDA recall events naming glass reached 11 in 2026 by 16 November, against 4 to 6 a year in 2023 to 2025.
- Scope ran from one lot code (3,585 quarts of ice cream) to 29 best-by dates and 2,209,728 tubs. The plant's codes and check records set the size, not the object.
- Example plant: a failed metal detector test can hold up to 2,700 half-gallons for each hour since the last passed test, $15,757 to $16,286 at US shelf prices. A recall by one best-by date puts $252,115 to $260,582 at stake.
Tracker, updated 16 November 2026. Next update: December 2026. Sources are listed at the end.
Every month, food regulators publish short notices for product that may contain something hard, sharp or simply out of place. Read one at a time, they say little. Read together, they show a pattern a plant can act on: the object decides whether there is a recall, and the plant's own records decide how big it gets. This page collects the notices, reads the codes and quantities in them, and turns them into checks for a food or beverage line. The record that narrows a hold is the same minute-by-minute line record that OEE for food and beverage is built on, so the two jobs share one data source.
New cases go at the top of the table each month. Older cases stay, so the page builds into a reference of what widened or narrowed real recalls.
The cases: foreign material recalls, August to November 2026
The table lists every foreign material notice we could read on a regulator's own page for the period: FDA Enforcement Report records with report dates from 5 August to 28 October 2026, notices from the Government of Canada's recall site for October and November, and notices from Food Standards Australia New Zealand (FSANZ) for October. Where one firm recalled several products for one cause, they share a row. "Scope as published" is what the notice itself lists, nothing more.
| Notice | Firm and product | Material | Class | Scope as published | Plant lesson |
|---|---|---|---|---|---|
| 2 Nov 2026, Canada | Molson Canada 2005: 29 beer products, 7 brands | Pieces of metal | Class 3 | 29 products, 21 codes, national | 21 codes spread over 29 products. Count how many products one run of your filler can end up in, before an incident rather than during one. |
| 30 Oct 2026, FSANZ | Maria's Pasta: 6 Bolognese sauce and pasta products | Metal | Not classed by FSANZ | Use-by 30 Oct to 4 Nov 2026, Victoria and online | All six carry Bolognese in the name. When one component feeds several products, trace the component batch first, then the SKUs. |
| 28 Oct 2026, FDA | Gearharts Fine Chocolates: boxed chocolates, 4 oz and 8 oz | Broken glass, found in a finished piece | Class II | 4,703 pieces, best by 1 to 21 Oct 2026, 17 states | Three weeks of best-by codes in scope. A glass breakage log with times is what cuts a glass event from weeks to hours. |
| 24 Oct 2026, FSANZ | Arla Foods Mayer Australia: protein puddings, 2 flavours, 200 g | Glass | Not classed by FSANZ | One best-before date per flavour, NSW, ACT and QLD | One date per flavour kept each recall to one date code. A date code that changes every day does that work for nothing. |
| 21 Oct 2026, FDA (posted 2 Oct) | Stewart's Shops: 2 half-gallon ice cream flavours | Metal object, found by a consumer | Class II | 21,152 cartons, one best-by date each, sold 17 Jul to 2 Oct, NY, VT, NH | Recalled 11 weeks after the first sale, once a consumer found the metal, on a 7-month shelf life. A catch at the line is the only cheap catch. |
| 14 Oct 2026, FDA | G L Mezzetta: peperoncini, 32 fl oz jars | Mice (pest) | Class II | 7,497 cases of 6, one lot (720106), 22 states | A pest is not a line foreign body: storage and exclusion checks are the control. One lot code kept it to one lot. |
| 6 Oct 2026, Canada | Loblaw: President's Choice English-Style Gingersnap Cookies, 350 g | Metal wire | Class 2 | 2 best-before dates, 34 days apart, national | Two dates five weeks apart. Ask which wire parts both runs shared: sieves, screens, brushes, belt lacing. |
| 30 Sep 2026, FDA | WWF Operating Company: So Delicious cashewmilk frozen dessert, 16 oz | Small stones and other hard objects | Class II | 2,209,728 tubs, 29 best-by dates from 28 Aug 2027 to 3 Apr 2028, nationwide | Stones are not machine wear. Treat the raw material lot as the first suspect and trace it across every run it fed. |
| 9 Sep 2026, FDA | Ukrop's Homestyle Foods: bread pudding with vanilla sauce, 13.8 oz and 4 lb | Metal fragments | Class II | 10,487 units, best by 6 Jul to 3 Aug 2026, 4 states | 29 days of codes for one recipe. Without a proven last good check, the window runs back weeks. |
| 2 Sep 2026, FDA | Dreyer's Grand Ice Cream: Outshine fruit bars, 6 products | Glass pieces | Class II | 1,974 pallets, 54 batch codes, best before May to Nov 2027, 32 states plus PR and VI | Six batch codes appear on two flavours each. Map every batch code to the line and hours it covers before you need to. |
| 26 Aug 2026, FDA | Gifford's Dairy: fudge ice cream, 1 quart | Rubber pieces | Class II | 3,585 units, one lot (26184), MA, ME, VT | One lot code held it to 3,585 quarts. Rubber is not what a metal detector is set to find, so seal and gasket checks at each wash are the control. |
| 26 Aug 2026, FDA | Grand Central Bakery: potato sourdough loaves, 3 sizes | Metal pieces | Class II | 517 loaves, no codes, sold on 10 Aug 2026, WA only | No codes at all, yet only 517 loaves: a one-day sale and a short shelf life did the work a code would. |
| 12 Aug 2026, FDA | PT Organics: banana and strawberry puree, 4 oz pouch | Thin, soft plastic strands | Class II | 4,385 pouches from packaging lane 4, 6 best-by dates | The lane number printed on the pouch seam put one lane in scope, not every lane that ran those days. |
| 12 Aug 2026, FDA | Cheer Pack: spouted foil pouches (part CP6088) made for that puree | Pouch may release thin plastic strips | Class II | 322,450 pouches made 25 to 31 Oct 2025, 12 lots | The pouches were made about nine months before the recall began. Log each packaging lot against the runs that used it. |
| 5 Aug 2026, FDA | Galliker Dairy: chocolate reduced fat ice cream mix, 5 gal | Metal shavings | Class II | 74 cases, one code (JUN1226), 5 states; closed | Shavings suggest a cutting or wear source upstream of filling. One code kept it to 74 cases. |
Notice pages read for the table: the FDA's Stewart's Shops posting, the Government of Canada notices for Molson (RA-82913) and Loblaw (RA-82751), and the FSANZ notice for Arla's protein puddings. FDA rows use the Enforcement Report fields for quantity, codes and distribution. We left out notices we could not open on the regulator's own site, including USDA meat and poultry notices and the UK notice reported for the same Arla puddings.
What the 15 notices say together
Three things stand out. First, the object is ordinary: metal in seven cases, glass in three, plastic in two, then rubber, stones and a pest. Second, only two notices say how the problem was found. Stewart's says a consumer found a metal object in a carton; Gearharts says broken glass was found in a piece of finished product. The other thirteen are silent, which means a plant learns little about root cause from other people's recalls and a lot from the records kept on the plant floor. Third, the size of each recall had almost nothing to do with the object. It followed the codes and records.
The long view points the same way. The CFIA's recall incident statistics count 17 extraneous material incidents in fiscal 2021-22 and 37 in 2025-26, when that cause made up 20.0% of all 185 incidents. In the US, our count of FDA Enforcement Report events whose reason names a foreign material moved between 9% and 12% of all food recall events a year. The method is a keyword search of the reason text, so treat the counts as close rather than exact.
| Year | FDA food recall events | Naming a foreign material | Share | Naming glass |
|---|---|---|---|---|
| 2023 | 480 | 54 | 11.3% | 6 |
| 2024 | 461 | 42 | 9.1% | 5 |
| 2025 | 575 | 70 | 12.2% | 4 |
| 2026, to 16 November | 427 | 52 | 12.2% | 11 |
Source: our count of FDA Enforcement Report data by report date. An event is one recall, which can cover many products. "Naming a foreign material" means the reason text contains metal, glass, plastic, rubber, wood, stone, rock, foreign object, foreign material, foreign matter or extraneous.
Glass is the number to watch. Eleven FDA events named glass in 2026 so far, against four to six a year before. Two of this period's three glass cases spread over weeks or months of codes, which is what glass does when nobody can say when the breakage happened.
What set the size of each recall
| What set the scope | Cases in the tracker | Scope range |
|---|---|---|
| One lot or one code | Gifford's Dairy, Galliker Dairy, G L Mezzetta | 74 cases to 7,497 cases |
| One date code per product | Stewart's Shops, Arla | One best-by date per flavour; 21,152 cartons at Stewart's |
| A sale date, no code | Grand Central Bakery | 517 loaves |
| A packing lane mark | PT Organics | 4,385 pouches from lane 4 |
| Two separate dates | Loblaw | 2 best-before dates, 34 days apart |
| Weeks of codes | Gearharts (21 days), Ukrop's (29 days) | 4,703 pieces; 10,487 units |
| Months of codes | Dreyer's (best before May to Nov 2027), WWF (29 dates over about 7 months) | 1,974 pallets; 2,209,728 tubs |
| A shared component or input | Maria's Pasta (6 products), Molson (29 products), Cheer Pack (pouch lots) | 6 to 29 products; 322,450 pouches |
The pattern is plain. Where a plant could point at one lot, one date or one lane, the recall stayed inside that lot, date or lane, whatever its size: 74 cases at Galliker, 3,585 quarts at Gifford's, 7,497 cases at Mezzetta, 21,152 cartons at Stewart's. Where it could not say when the problem started, the window ran back to the last moment it could prove things were fine, and that was weeks or months earlier.
Where foreign material gets into a line
Each material points to a different part of the plant, and each needs its own control. A metal detector covers one row of this table. Europe sees the same sources: the European Commission's Alert and Cooperation Network report for 2025 says foreign body cases in processed meat products, such as metal fragments and plastic pieces from the manufacturing process or from packaging materials, "remain common".
| Material | Where to look first | What finds or prevents it | Cases here |
|---|---|---|---|
| Metal pieces, fragments, shavings, wire | Cutting parts, sieves and screens, worn fixings, brushes, tools and parts left after maintenance | Metal detector with timed tests, magnets, wear part checks | Stewart's, Ukrop's, Galliker, Loblaw, Maria's, Grand Central, Molson |
| Glass | Lights, gauges, sight glasses, glass containers near open product | Glass and brittle plastic register, a breakage log with times, X-ray where fitted | Gearharts, Dreyer's, Arla |
| Plastic | Film and pouch material, scoops, belts, guards | Packaging lot control, checks at changeover, supplier records | PT Organics, Cheer Pack |
| Rubber | Seals, gaskets, scraper blades | Seal and gasket checks at each wash, with the time recorded | Gifford's |
| Stones | Raw material | Supplier checks, sieving or sorting of the raw material | WWF |
| Pests | Storage, raw material, the building | Pest control and storage checks | G L Mezzetta |
Size matters as much as material. The FDA's compliance policy guide CPG Sec. 555.425 treats a hard or sharp object of 7 mm to 25 mm in ready-to-eat or minimally prepared food as grounds for action, and an object over 25 mm when an FDA health hazard evaluation finds a hazard. Objects under 7 mm can also count when the food is meant for a special-risk group. That last clause matters for one row above: the PT Organics puree is sold for children from 6 months old, according to its FDA record.
The check interval decides the hold
When a metal detector fails its test, the question is not only what went wrong. It is how much product ran since the last test that passed, because all of it is now suspect. US preventive controls rules put this in writing. The corrective action section, 21 CFR 117.150, requires that all affected food is evaluated for safety and prevented from entering commerce. The last passed test is where "affected" begins.
So the interval between tests is a scope decision, not only a quality schedule. Two records turn an interval into a number of units: the time of each passed test, and the line count between tests. The first sits in the detector log or the quality record. The second sits in the line counter, which is the same count that feeds availability and performance in OEE. When both exist, "since the last good check" becomes "4,812 cartons, pallets 3 to 7". When either is missing, it becomes "since start-up" or "since yesterday".
The same logic applies to the other controls. A sieve found torn at its 14:00 check puts everything since the 10:00 check in doubt. A gasket found split at a wash puts the run before it in doubt. What keeps the hold small is the time on each check, kept where the production record can line up with it.
Worked example: a half-gallon ice cream line
Ice cream and frozen desserts make up five of the eleven FDA events in the tracker, so the example uses that product. Every operating figure is Example data; the price is a published statistic.
- Example plant: one line filling half-gallon cartons at 45 a minute, 2,700 an hour, 16 production hours a day over two 8-hour shifts.
- Detector tests: at start-up, after each changeover, and at a fixed interval. Each test pauses the line for 2 minutes.
- Price: the BLS US city average price for prepackaged regular ice cream was $5.836 per half gallon in July 2026 and $6.032 in October 2026. That is a shelf price. A plant's selling price is lower, so read the money below as the top of the range at stake, not a loss figure.
Step 1: the hold after a failed test. In the worst case the test fails just before the next one is due, so the hold is a full interval of output.
- Every hour: 2,700 cartons. 2,700 x 5.836 = $15,757.20; 2,700 x 6.032 = $16,286.40.
- Every 2 hours: 5,400 cartons. 5,400 x 5.836 = $31,514.40; 5,400 x 6.032 = $32,572.80.
- Every 4 hours: 10,800 cartons. 10,800 x 5.836 = $63,028.80; 10,800 x 6.032 = $65,145.60.
Step 2: the recall if nothing is caught at the line. If the first sign is a consumer finding metal, as at Stewart's, the plant recalls by code. With a best-by date that covers one production day, that is 16 hours of output: 16 x 2,700 = 43,200 cartons. 43,200 x 5.836 = $252,115.20; 43,200 x 6.032 = $260,582.40. That is 16 times the hourly hold, and the product has already left the building.
Step 3: what the tests cost in line time. Held product is not lost by default. It can be re-run through a working detector and released if clean, so the bigger daily cost is line time.
- Hourly tests: 16 a day x 2 minutes = 32 minutes, 3.3% of the 960-minute production day (32 / 960).
- Four-hourly tests: 4 a day x 2 minutes = 8 minutes, 0.8% of the day (8 / 960).
- The difference is 24 minutes a day, which is 24 x 45 = 1,080 cartons of output the line does not make.
Step 4: the re-screen after a failure. Re-running a hold at line speed takes as long as it took to make it: up to 1 hour of line time with hourly tests, up to 4 hours with four-hourly tests. The gap is 3 hours, or 180 minutes.
Step 5: the break-even. Hourly testing costs 24 more minutes a day. On the day a test fails it saves up to 180 minutes, and about 90 on average, because a failure lands on average halfway through the interval (30 minutes of re-screen against 120). 90 / 24 = 3.75. If the detector fails a test more often than once in about 3.75 production days, hourly testing saves line time on top of shrinking the hold; on the worst-case gap of 180 minutes the break-even is 7.5 days. If it fails far less often, the longer interval costs less time, until the day a failure means scrapping product instead of re-screening it. At that point the money in step 1 applies, and the gap between $15,757 and $65,146 is the price of the longer interval.
The example turns on two numbers a plant has to measure for itself: how often its tests fail, and how long a test really pauses the line. Both come from logging every test with its time and result. If test packs can run through without stopping the line, step 3 shrinks to almost nothing and short intervals are close to free.
Book a demo and see losses per line and shift.
Choosing a test interval: a decision table
The rules read for this page, 21 CFR 117.150 and the FDA's foreign object guide, say what to do with affected food. They do not set a test interval; the plant's own hazard analysis does. This table uses the Example plant's numbers to show the trade.
| Interval | Test time a day | Worst-case hold | Worst-case re-screen | Fits a plant that |
|---|---|---|---|---|
| Every 30 minutes | 32 tests, 64 min | 1,350 cartons ($7,879 to $8,143) | 30 min | packs food for infants or another special-risk group, or had a failed test in the last month |
| Every hour | 16 tests, 32 min | 2,700 cartons ($15,757 to $16,286) | 1 h | runs a long shelf life product in high volume, where a recall by date code is costly |
| Every 2 hours | 8 tests, 16 min | 5,400 cartons ($31,514 to $32,573) | 2 h | has months of logged tests with no failures and runs a short shelf life product |
| Every 4 hours | 4 tests, 8 min | 10,800 cartons ($63,029 to $65,146) | 4 h | pauses the line for each test, has a clean logged failure history and can re-screen on a second detector |
Questions to settle with the detector supplier and your own team before picking a row:
- Can the test pack run through without stopping the line?
- Is every test logged automatically with its time and result, or written on paper at the end of the hour?
- Does a failed test stop the line, or only raise an alarm someone has to notice?
- Does the reject device confirm each reject, and what happens when the reject bin is full?
- Is there a second detector where held product can be re-screened without taking the line?
What each manager takes from this month's cases
Plant manager
Ask for the scope number before an incident, not after it. Suppose the detector failed its 14:00 test today: how many units are on hold, under which codes, on which pallets? The answer sorts your plant into one of the rows of the scope table above. Ukrop's recall ran across 29 days of codes; Gifford's stopped at one lot. The difference is records, and records are a budget line. Weigh the cost of a code that changes by line and hour against a window that runs back weeks.
Production manager
Tests, rejects and re-screens are line time, and they belong in OEE with honest reason codes rather than hidden inside "minor stops". Classing them is covered in food safety stops in OEE, and the reject side in checkweigher rejects in OEE. Keep the line counter running through every test so the count between two tests is never a guess. And take the Stewart's lesson to the shift briefing: the recall came 11 weeks after the first sale, once a consumer had found the metal object, on a product with a 7-month shelf life. The line is the only place a catch is cheap.
Quality and maintenance manager
Six of this period's 15 cases name a material that worn or broken plant parts can shed: rubber at Gifford's, shavings at Galliker, wire at Loblaw, glass at Gearharts, Dreyer's and Arla. The notices do not name a cause; only Gearharts' says the glass was broken. Put seals, gaskets, scraper blades, sieves, screens and brushes on a timed check list, and record the time of each check, not just a tick for the shift. Keep the glass register's breakage log with times too. For inputs, the Cheer Pack and WWF cases say the same thing from two sides: log every packaging and ingredient lot against the start and end time of the runs that used it.
This quarter: a foreign material checklist
- Decode every code format your plant prints on one page: what each character means, and whether it carries line, date, shift or lane.
- Pick one line. From last week's log, not the plan, write down the real interval between passed detector tests on each shift.
- Multiply each interval by the line's output per hour from the counter. That is your worst-case hold today.
- If one line feeds several lanes or heads, add a lane mark to the pack, as the PT Organics pouch seam did.
- Log each packaging and ingredient lot against run start and end times, in a structure like a lot traceability data model.
- Put wear parts (seals, gaskets, scrapers, sieves, screens, brushes) on a timed check with the time recorded.
- Keep a glass and brittle plastic register with a timed breakage log.
- Run one drill: pick a time from last month, pretend the detector failed its test then, and time how long it takes to name the units, codes and pallets in scope.
Counting what ran between two checks
A hold is a count with two time stamps. The check log gives the time stamps; the line gives the count. When the line count is kept by minute and shift, a quality manager can answer "how much since the last passed test" in seconds, and a plant manager can see what tests and re-screens cost in availability over a month.
Fabrico is a cloud OEE and light MES platform. It shows OEE and losses by machine, line and shift, from the PLC or a Fabrico IoT device, with reasons operators add on a tablet or phone. On manual stations, AI cameras (an add-on) count output. See how it works in OEE software, light MES and OEE with computer vision.
Book a demo and see downtime per line and shift.
What people ask next
What is a foreign material recall?
A recall of food that may contain an object that should not be in it, such as metal, glass, plastic, rubber, stones or wood. Canada files these under extraneous material and the FDA under foreign objects. In this tracker, 15 notices from August to mid-November 2026 covered metal (7), glass (3), plastic (2), and rubber, stones and a pest (1 each).
What size of foreign object makes food adulterated in the US?
The FDA's compliance policy guide CPG Sec. 555.425 treats a hard or sharp object of 7 mm to 25 mm in ready-to-eat or minimally prepared food as grounds for action, and an object over 25 mm when an FDA health hazard evaluation finds a hazard. Objects under 7 mm can also count when the food is meant for a special-risk group, such as infants.
What should a plant do with product after a failed metal detector test?
Treat everything made since the last passed test as affected: hold it, evaluate it and keep it out of commerce, as the corrective action rule in 21 CFR 117.150 requires for affected food. Re-screening through a working detector is one way to evaluate it; your food safety plan sets the method and who signs off the release.
How often should a metal detector be tested?
The rules read for this page do not set one interval; the plant's hazard analysis does. The interval sets the worst-case hold: on a line making 2,700 units an hour, hourly tests cap a hold at 2,700 units and four-hourly tests at 10,800. Log every test with its time and result so the boundary can be proved.
Why do foreign material recalls cover so many products?
Because the products share something: a sauce, a liquid, a pouch lot or a raw material. In this tracker, Maria's Pasta recalled 6 Bolognese products, Molson Canada 29 beer products, and a pouch maker 322,450 pouches made for one puree.
How does OEE data help after a foreign material incident?
OEE software records output and stops by line and time. That turns "since the last passed test" into a number of units and shows how much line time tests, rejects and re-screens take each month. It does not replace a metal detector, X-ray or a food safety plan.
Sources
Sources accessed on 16 November 2026 unless marked as accessed on another date. Dates are the publication dates shown on each page.
- Canadian Food Inspection Agency, Statistics: food recall incidents and food recalls, 13 August 2026, accessed 16 November 2026
- U.S. Food and Drug Administration, Enforcement Report data for food (openFDA), foreign material records, report dates 5 August to 28 October 2026, accessed 16 November 2026
- U.S. Food and Drug Administration, Stewart's Shop Corp. recalls Heavenly Medley and Vanilla-Chocolate-Strawberry due to possible foreign object, 2 October 2026, accessed 16 November 2026
- U.S. Food and Drug Administration, CPG Sec. 555.425 Foods, adulteration involving hard or sharp foreign objects, revised May 2005, accessed 16 November 2026
- Government of Canada, Recalls and safety alerts: Various brands of alcoholic and non-alcoholic beer recalled due to pieces of metal (RA-82913), 2 November 2026, accessed 16 November 2026
- Government of Canada, Recalls and safety alerts: President's Choice brand English-Style Gingersnap Cookies recalled due to metal wire (RA-82751), 6 October 2026, accessed 16 November 2026
- Food Standards Australia New Zealand, Arla Protein Salted Caramel and Chocolate Flavour Pudding 200g recall, 24 October 2026, accessed 16 November 2026
- Food Standards Australia New Zealand, Maria's Pasta various products recall, 30 October 2026, accessed 16 November 2026
- U.S. Bureau of Labor Statistics via FRED, Average price: ice cream, prepackaged, bulk, regular, per half gallon, U.S. city average (APU0000710411), October 2026 data, accessed 16 November 2026
- Legal Information Institute, Cornell Law School, 21 CFR 117.150 Corrective actions and corrections, accessed 16 November 2026
- European Commission, Alert and Cooperation Network 2025 annual report, 2026, accessed 16 November 2026
Related articles

Real Fabrico tiles, sample data.








