Dear Charles,
Sorry for posting my opinion so late in the discussion, I am usually working on the mode of a turtle.
From where I come from, we always say "Never argue with an auditor".Generally, most companies in the food industries today have Metal Detectors in their process lines designed for the purpose of detecting metal hazards. Often this step of the process is considered significantly risky to be classified as a CCP (Critical Control Point) - most food auditors insist it is anyway.


I would not agree that Metal Detecting to be validation, if it is not a CCP, it may well be verification, to verify that the grinding process is well controlled, not introducing any metal fragments.
IMO. I would consider metal detecting to be a CCP, irregardless whether it is a process step. Say we ommit Metal Detecting as a process step from the flow diagram. The hazard of metal fragment would generally come from the process step of grinding, whereby, when we look into the Codex decision tree, answer for Q2 would be No, Q3 might be Yes, while Q4 would be No, rendering the process step of "Grinding" as a CCP. How would one monitor that? By using a metal detector. So, at the end, even if "Grinding" is a CCP, monitoring would still round back to metal detecting.
IMO, I'd say the last point, be it before bagging or before bulk loading. Same concept as the many filters in an oilmill.Indeed a metal detector is a CCP regardless. Interestingly, in a cocoa powder manufacturing facility where metal detectors are all over the entire piping line, where should the CCP be?
Say, would there be so many metal detectors in the cocoa powder plant, or is it just magnets?? Controlling magnets and controlling metal detector are very different.
Metal detectors are usually simpler to control, IMO, because most metal detectors work in a way that it detects every single product and reject those with metal detected in the product, so the affected products would be quarantined.
Whereas for magnets, when we actually catch something on the magnet, it may not mean that all metals were caught. We may have to think not just about the gauss, but also the flow-rate of the product as a CL.
To me, O' PRP and CCPs are similar. We choose to "elevate" certain O' PRP to be CCP to further emphasize on the control point, because it has a direct impact on the safety of the product. In time, the CCP if well controlled might even be classified to be an O' PRP again, because the likelihood of the hazard would change from high to low.I know - it sounds pretty
but lets be realistic. We have all the OPRPs in place and we still have to "create" a CCP for a monitoring step to a CCP.
I have seen an instant noodle processing plant having frying as a CCP initially, and the frying temperature as monitoring. But after a few years, they noted that the fluctuation in the frying temperture was minimal, but the process step which was creating the problem of "under frying" was actually the weighing and the cutting of the noodle strands - too high volume in a contrained space affects the frying process, so "weighing" became the new CCP.
Got another one for ALL to think about - A fish canning facility with a vacuum seamers when seaming is considered a CCP step (torque on seaming integrity etc).........then immediately after seaming, we have the seaming detector (again this is usually treated as a CCP when ALL that it does is only a validation step to the identified CCP.
Surely we cannot have two CCPs when one is good enough. If we look at the usual flow of determination under the Decision Tree...........it strictly mentions "process step" and NOT Monitoring step.
Never seen a seaming detector, not sure how it works, but in canning, what some people do is just to cut open the seam and check on the over-lap, maybe it is the "old" conventional way, but still works. I'd reckon the seaming detector functions on the same concept.
I agree that it shoud be either the torque or the seaming detecting to be a CCP. However, I do believe, that the seaming detector is again, verification, to verify that seaming is carried out properly.
Validation would be carried out initially on the CL of the torque: is that particular value sufficient to ensure proper seaming?
IMO, validation is not something that is carried out on routine basis (as in seaming detecting or metal detecting), it is usually carried when we first design ordevelop the system, or when we modify the system.
Kindly let me know if my concept is wrong.

Cheers,