I'm certainly not an expert on all those troubles, but I can tell you that the grid stays on during my outputs as well, and I've done several prints over the last few weeks, so it seems to be normal.
Looking at your pic, if that was my set-up, I would have put more supports on it. It seems like a pretty heavy model for the four supports. Every time I have a failure like you describe, I rebuild the layout with more supports, and it's helped every time.
Also, seeing how you said you ended up getting disks floating around, with no real build happening, it seems we can conlude that the support bases didn't adhere to the build table as well as they need to - make sure you get your build table flat on the surface of the VAT window. I think it's somewhere in the directions, where with the screws loosened, you send the command to the B9 to calibrate the table to the PDMS vat window surface with a "g" command. Not sure about the order of things but you probably know what I'm talking about.
With the four screws loose, and the build platform raised and above the higher window build side of the vat, send the machine the "g" command and it will lower the build table until it stops. Then gently push the build table flat on the VAT window surface. Tighten the four side screws. It's recommended that the top build-table screw be removed at this point, I'm guessing because by pressing the table against the PDMS like this, it can create a slight surface suction that can cause the table to damage the PDMS if you just command the machine to raise it. I may be wrong here, but I've had good luck removing the build table's top hold screw as directed in that wiki instruction page, and raising the z-axis manually with the right Z switch, lifting the arm away from the build table and leaving the build table apparatus flat on the surface of the vat. Then I lift the build platform out by tilting it at an angle, to gently separate it from the PDMS.
It's important to note that you'll probably need to do this every time you change vats or after a new PDMS coating, since the surface is never going to be exactly the same on any two vats coated like this. So a set-up that might have been flush and flat on one vat may actually be askew from the surface of another vat.
ALSO, I don't remember who it was, but someone here has a method where they gently press down on the build table's arm during the first layer of a build, so that first layer really gets the best chance it can to adhere to the build table. I've done that a bit after having some failures like what you describe. I start the print, then press lightly on the sides of the black u-bracket of the build table for about 20-30 seconds during the first layer's exposure, and then stop when I think it's about to finish. It's tricky because you want to remove any pressure before the layer finishes and the vat moves. Also, I think too much pressure during this could flex the Z-worm gear shaft more than it needs to, so use care if you opt to try it.