Nope, the message tells you exactly what you need to do - take target_z down a bit - I think I have mine set at -0.75. And it literally tells you what the problem is - the initial non-contact zeroing is high enough that your probe thinks your hotend is at -0.45mm and it hasn’t actually touched the bed yet. You can look up the command on the eddy-ng wiki pages to do a tap homing outside of your print start macro to more easily diagnose - get down nice and low so you can see how close the tip is getting to the bed.
I can almost guarantee that your problem could only possibly be one of your bullet points - eddy positioning. I’m using vlad’s own mount.
Here’s why it’s none of the others. It’s not cable noise - that would just cause klipper to hard-stop because it can’t communicate with the mcu. It’s not the screws, because they’re stationary with respect to the eddy - they’re also 20mm away from the coil. Bad usb connection on the hub would also lead to mcu faults hard-stopping Klipper. Bad tension would only be relevant if you’ve been dicking around with your z-axis belts, and it would be less random.