It's an intentional reed switch or magnetometer built into the device so that it can be put into a Fail-Safe mode by a paramedic or doctor. Transiently tripping the circuit occasionally likely wouldn't hurt anyone. The fear would be if someone fell asleep with a magnet on top of their device or something, and left the device in its Fail-Safe state for an extended period of time.

It's an intentional feature. Implanted devices by design go into a temporary Fail-Safe mode of operation in the presence of a magnetic field, so that paramedics and doctors can quickly put a patient's device into a sane state when they suspect it's malfunctioning.

I have a pacemaker, and it uses BLE. It broadcasts announcement packets once a minute, and claims to have a 10 year batter life. It automatically shakes hands with a cell modem base station on my nightstand multiple times a day. Modern pacemaker's aren't "susceptible" to magnetic interference or going to forget their programming by someone holding an IPhone too close. They are intentionally designed to go into a specific Fail-Safe mode of operation in the presence of a magnetic field. The purpose is for paramedics and doctors to be able to quickly put them into a known state should they suspect it's misbehaving, just by putting a magnet on someone's chest.

I actually tried that, but I had already finished the compact 3D printed enclosure, and there wasn't much room left. I had a digispark on hand and was going to deadbug the 8 pin AVR from it in the available space, but midway through debugging the firmware I blew it up. So that's when I gave I2S a try and got good enough results.

I recently based a hobby project on an ESP32. I recently had to get a pacemaker and I made a personal EKG holter monitor that logs to SD with real-time display over wifi. SD was straight forward. Wifi was straight forward. Creating tasks was as straight forward as FreeRTOS makes it. A lot of example code is exceptionally crufty and would be confusing to someone with a lot of experience. The internal ADC is a steaming pile. In software triggered mode, it intermittently acts as if the external pin isn't electrically connected to the sampling circuit. I was trying to sample very large magnitude 350uS pulses (my pacemaker impulses) sampling at 8Khz, and it would completely miss them half the time. It isn't a bandwidth issue. Sampling a function generated waveform would just look nasty too, regardless of frequency. I eventually got the poorly documented I2S sample mode running. It doesn't use the same "Real Time Controller" middleman. It picks up all the 350uS pulses reliably now and is "good enough" for my purpose, but it's still rather ugly. I have the I2S oversampling 5x at 40Khz and averaging in software and it looks equally bad, so it might just have laughably bad non-linearity. I probably spent 30 hours battling that ADC. It is really too bad, because otherwise the ESP32 seems like a great chip.