1. As a standard we tie 24V negative to ground to set the ground potential. This assists with troubleshooting whereas you can just place your black multimeter probe on any metal part and take readings. However this seems to defeat the purpose of a galvanically isolated SMPS. I'm thinking of a scenario where someone accidentally touches 24V to a conduit or enclosure and now you have a path back to 24V- through the building ground. Not a huge dead because most likely there is like a 2Amp fast acting fuse that will blow. But if it was a truely floating power supply this prevent this, right? It can introduce potential ground loop issues on analog control signals which are a nightmare to troubleshoot. I'm thinking that at most one should tie all of the DC power supplies in a panel together to set the same reference, but only if you REALLY need to for some reason.
It does assist with fault finding, but there are other reasons behind this. I am going to take it you are talking about some sort of I/O panel here. In most cases your I/O cards would be on some sort of back plane which would have it's own power supply. You would have a second power supply (load power supply) which would then drive some other devices which need the extra power. Now depending on the "class" of power supply you have, the secondary side may or may not be tied down to the primary earth. You may have multiple analog signals that you are measuring in the form of 4-20mA etc. Those signals would be run in a twisted pair cable which in turn would have an external screen. The screen would be open ended on one side (usually the transmitter) and tied to a drain (earth) on the measurement side. This is to get rid of all the noise the signal may pick up from the field installation. Now lets go back to the two power supplies. Lets say that the two supplies are not at the same potential. This would interfere with how your card measures the signal (depending on the card type). It gets worse when you have a mixture of source and sync on analogue. In order to ensure that we are always at the same potential - we tie the secondary (0VDC) to the same reference - in this case earth. That way - no ground loops and no problems for the maintenance personnel.
Now this is said assuming that you are not working in a hazardous (explosive) type environment. In the good old bad days, we used to deal with the hazardous environment issue by using a thing called a Zener barrier. The purpose of this device was to limit the energy transferred into the hazardous area for the signal measurement to such an extent that you would not be able to ignite any potentially explosive gasses around even if you were to short the two signal wires out and make them spark. The issue with this design, was that the barrier needed a place to dump this extra energy to. This is what is known as an intrinsic safe earth. Now this is a really big pain in the but to maintain and is just a headache. As you can imagine, the impedance of this earth is extremely low. To overcome this, some bright sparks invented the isolator. It works the same, except the safe side and hazardous side are galvanically isolated. But again, everything is at the same zero volt reference and we don't require the intrinsic safe earth.
Somehow, this was never communicated to the younger guys and they just continued to design in the same manner without understanding why they were putting things in. You will see this often in instrument loop drawings where they might mention a thing called HQE (High quality earth) where they will tie the cable screen to and then hopefully tie that back to the main earth. So HQE and main earth are the same thing. You can find quite a lot on this if you look at the ISA website.
2. We don't fuse the output of a 24V power supply with a fuse rated to its max current. Power is immediately distributed into a row of fuses that then go out to power individual devices requiring 24V power. We will always size the power supply so that it won't ever exceed its rating. However it seems like a miss to NOT immediately fuse the output of a 10A power supply with a 12A fuse(or something), THEN distribute the power to the individual branch circuits. This is what we do with the incoming 120V AC which typically powers the panel. We first go into a circuit breaker, then immediately go into fuses for each 120V branch circuit. Why wouldn't we do the same for 24V power?
Why would you need to have a main fuse over the bank? The circuit breaker you refer to is only there to isolate the power - it is not there for protection. The fuses are the protection. Again, go look at many of the circuit breakers installed in your panels and then look at the tripping curve of the breaker - it will never protect anything in the panel, it is way too slow.
3. I don't really see the point in having an e-stop reset momentary pushbutton. I mean the ESTOP button is already a push-pull style button, requiring human interaction to pull it out which is, why make someone pull it out, then push a reset button?
It is hard to say why you have a reset over a latched e-stop. Again, you need to look at the process control philosophy. Perhaps you would not like some machinery to start up automatically once you release the e-stop, perhaps you would like to do an inspection first before you return the circuit to service? It is hard to say without knowing the philosophy or the environment. It may even be a SIL requirement for your specific plant?
Cheers