Because google is treating my comments as spam and I can’t respond to his post directly, here is my reply to @Alex_Skoruppa 's post. (People! Check your posts for 'comments marked as spam! Google is overly aggressive sometimes)
I’d suggest also looking at hardened balls in Delrin(Prototype is done in ABS) held in place by springs, rather than magnets. Steel on low frictions plastic and by changing the spring tention, you can control how much weight your head has. Plus you can keep the weight down by not having to have two lots of metal at either end.
We didn’t like the wear of steel on the soft nickel coating of the magnets, nor when the magnets let go, the head crashing into the bed.
This way, when it fails, is a bit spectacular, but the head ends up dangling on the retracted springs above the bed, rather than falling(In this case some 700mm) and crashing into the expensive glass bed.
This was our biggest problem with magnets as a ‘commercial’ printer producer at DeltaMaker. It’s a huge liability issue having a 230C hot end, dangling over plastic, and magnets allowing things to release and the hot end going who-knows-where. Especially given that our customers (on the regular) run gcode not designed for the machine, having it go cockeyed in all sorts of funny ways. I like the idea behind magnets, and I’ve toyed with it on my own personal printer – but it’s definitely not a solution that everyone can use.
@ThantiK If your joints are conductive (metal-to-metal, some lubricants that would interfere with the conductivity), you could run a low-voltage signal though all of the joints, up one side and down the other so that they’re all in series. If the circuit is broken, trigger an emergency stop and shut down. With the right logic, you could make your probe line serve double-duty for this, since you only need that when starting a print anyway.
@Whosa_whatsis , we contemplated this exact solution, we came to the conclusion that having your printer ever disassemble itself at all was going to make it seem as cheap as a rock’em-sock’em-robot.
@ThantiK Totally agree, any way you can drop a hot end is a giant red flag. It screams amateur hour – something like that should immediately come up in any formal risk assessing process, such as is required for CE certification.
Any industry that does serious lifting or rigging has had a handle on this for decades.
I kinda like the idea that the head will pop out of place instead of parts getting ripped up or bent out of spec if the g-code is wrong.
But yeah, you need some kind of safety harness and a stop for when the harness is employed.
The big advantage to ‘self distructting’ printer so far is if anything goes wrong, the machine won’t put any excessive force on anything delicate and sprong.
@paul_wallich this sort of does that, the springs/cords hold the head dangling in the air. it wouldn’t take too much to also have the arms secured by loose thread to each carriage too, so they don’t go flying.
@Whosa_whatsis and I like the idea of detecting arm fails. It would need to be implemented in the head, a simple current loop up each arm, across the carriage and back down again to the next pair, along with some conductive grease on each joint and it would work. A bit overkill for a hobby machine but combined with filament jam sensors might be useful on a more professional machine.
A ‘safe’ fail is better than snapping or breaking something permanent!
@Electra_Flarefire Alternate approach: size your motors properly, so the machine isn’t powerful enough to self-destruct.
@Ryan_Carlyle that means an underpowerd machine with slow print speeds and high motor currents(Thus runs hot).
The ‘self destruct’ is mostly when it goes out of bounds somehow or you try and start the printer with the previous job still on the bed. 
@Electra_Flarefire I’ve had tons of violent crashes on deltas that can go >300mm/s, and never had any meaningful damage. That’s the great thing about steppers: they stall harmlessly if they hit a hard stop. You just have to build the arm strength proportionate to the motors, or select motors proportionate to the arm strength.
If a machine will destroy itself in response to a predictable user error, it’s not designed right. That’s engineering 101.
@Ryan_Carlyle very hard to detect ‘things still on bed’. Unexpected tangles. Making stupid mistakes like that. And the out of bounds thing is something we are working on, but while we are learning, having the machine ‘safely’ dismantle it’self is very handy. It can’t bend any brackets, damage any 3d printed parts before the motors stall out.
I’ll also note this is our first Delta, so it’s all learning.
And (depending on the setup) steppers that come up against hard stops can do a fair amount of damage to beds, nozzles, end stops and the like. And the electrical spikes of that stalling can potentially propagate into the rest of the system (if wires are bundled together). Speaking as someone who got to replace a mainboard and inductive sensor for apparently that reason.