RR NDB;
The oscillator frequency changes near a metallic object.
Not just ferrous! And the memory type inside is insensitive to magnetic field.
Lets be careful with the word "metal". Not all metals are ferri/ferromagnetic, and looking at the properties of brass and aluminium/aluminum should be enough to demonstrate. An old trick when working on tuned circuits using threaded ferrite cores in the coils was to test them by not attempting to screw the often breakable ferrite core, but instead used a simple yet not destructive way of checking.
The method used a standard poly/plastic tuning tool with a small rod of iron on one end and brass on the other. Inserting the brass end raised the frequency and inserting the iron end lowered the frequency.
In the case of the CSMU we are looking for a lump of steel, and the frequency of a tuned detection loop will lower ("dip") once the inductive field is intersected by the lump of steel. Your bold red correction of what I originally stated, is IMO not warranted. Yes, I knew the memory was "flash" from the Honeywell specs, but I didn't set out to put the original poster "down".