Im not a massive fan of using standard connectors for non standard uses. Any company that makes a device that uses usb-c to charge a device, but with unchecked 12v in there needs locking up!!

Using standard connectors does of course have many advantages. They are generally cheap and easy to source, and normally have great connector density allowing you to have lots of connections in a small space.

When I originally had the idea for CL-32 i had though of using an M2 socket for the connection of additional functionality. The non standard usage always put me off, and I wanted to keep the components on the one side of the board to keep manufacturing costs down. This lead me to use the reverse entry headers that were on my design.

Looking for parts for the board made me realize that its not always clear which headers are designed for bottom entry, and my past experience with surface mount headers hasnt been great. They are only held onto the board by the strength of the bond between the copper and the pcb. If the usage of this device is going to be in an educational setting, then I can see headers being ripped off regularly

Magnets? How do they work?

One of the comments on the project suggested looking at magnetic connectors. These would solve the heavy handed user problem, and the polarization of the magnets should stop people wedging the board backwards.

There are 16 i/o pins, and 2 power pins, so we would need a larger connector.
 

Using something like this would possibly fit the bill, but they only come in thru hole mount, making a large portion of the add on board un-usable for adding functionality. So the hunt continued..

M2 popularity

Looking in the tech field, the M2 slot is used a few times by other devices (see sparkfun MicroMod and Sipeed Lichee RV) so its not un-common to see it in non standard settings. I had though I could try and sit as close to a standard spec as possible, with the location of power and data lines, so started looking.

After some digging it actually looked like an E key pin layout actually fits much better than i thought. 

The USB connection to the ESP could be piped into the connector theoretically meaning that an old E-key device that worked over USB (modem/gps/etc) could possibly work as long as it was happy wth 3.3v.

The I2C connection would allow the keyboard and RTC I2C connection to also be broken out into the connector, allowing similar connectivity as the USB, and also adding to the available i/o.

The UART connection could allow backup access to the ESP if the normal USB connectivity fails.

The PCI-e pins could then be used for the i/o, keeping things within the "standard" with the location if data pins, and it might even let some crazy people try to talk to PCI-e devices with the ESP.

So the actual pin layout should be safe if there was any crossover in card usage (CL-32 and laptop) but the biggest issue could be voltage. A laptop card is designed to work on 1.8v but my cards will be using 3.3v, so using a CL-32 card in a laptop would be fine, but a laptop card in a CL-32 might let the magic smoke out

Seems like a small enough risk for me...