Any glass bed is likely to be more accurate than a cheap aluminum plate, and better than a used aluminum plate. That is part of why so many use a glass plate. It's easy enough to change regardless.
I have had a bunch of glass beds...
Soda glass - (also called soda lime or green glass), this lasted me 9 months. It broke when I had a cooling fan positioned wrong and it cooled one half and not the other. It flexed, but never broke. I had a head crash take a chunk out and it survived. Accuracy on it was described as "good". I have no idea what that means. lol Large chunks when it breaks.
Cheap window glass - I used it for 4 months before I replaced the printer. I have heard people say it lasts quite a while, and at $5 or so a plate, who cares. Beware head crashes though. Expect large chunks when it breaks.
Borosilicate - (a.k.a. oven glass, or corning ware) I had one plate for 3 months, a head crash (minimal damage) and too tight of a bed clamp eventually cracked it in half. Seems it doesn't like impacts or stress, this is partly due to thickness (3mm). It's hard to find thicker pieces and it's not cheap as cheap as plate glass. Accuracy on it is considered the same as tempered. Expect very large chunks when it breaks.
On any thin glass, what it is clamped to will have a more dramatic effect on flatness than the glass itself.
The thick tempered glass on my big printer...
While there is no industry standard, the industry generally considers tempered to have an accuracy of .0005in (.01mm) tolerance across the surface. This glass is thick and heavy, but I have yet to have an issue with it. In fact, I prefer it to every other bed I have worked with. Head crash, it's more likely to break the printer. It's nice because it feels solid, as opposed to the flimsy glass on my other printers. When this breaks, it breaks into thousands of tiny pieces, however, it's stronger than other glass by a lot. Mine is heavy enough to break a foot if it falls on you. The weight actually adds stability to the printer and there is no way to warp it.
As for heat times and insulation...
My small printer has a kapton heater on it, it takes about 4 minutes to heat. When I added the double thick borosilicate, heat times only went up 25% and my temps held better ones heated.
Regarding even bed temps, it's a joke, no matter what you put for a heater, your temps will vary. I found Kapton heaters and pcb heaters to vary 15 degrees from one spot to another as they heated up. An aluminum plate with glass will heat the most evenly, however once that warps, it's worse than the rest. Personally, I like to use a Kapton heater, and give it time to settle. The very edges will still be cooler, but that's true on any surface. Because it holds the heat, it should remain more stable than others across the surface once heated. The bad side is that if the glass breaks, your heater is stuck to it, this is a mixed blessing, as while it offers no strength, it will hold the glass together when it breaks, giving an added level of safety.
I don't dare heat the big one, while the glass was tempered to 1000 degrees and I have had quotes for a heater ($30), but it would take an insane amount of heat. It would use almost as much power as my air conditioner while running. Which would also need to run.
I use hairspray on my heated bed, and blue tape on the big printer. Works for me.
Delta printers are great, they are fast, but can be fickle, and have less documentation. The calibration issues can be solved using a few new parts that are out which allow it to auto calibrate. It takes a few extra parts, but makes the delta easier to calibrate than a cartesian. Documentation is a big problem for deltas still, so you may want to try a cartesian first. If it was me, I would build a delta or H-bot (core-XY), but then, I like a challenge.
They all print similar quality, remember, they almost all use the same basic components. I have seen everything from a Printrbot Simple, to Ultimaker 2, to 1meter tall deltas make nearly equal prints, the difference was how they did it, how fast and for how long. The best consistent quality I have seen was from a well tuned Ultimaker and the fastest was a Griffin delta, which was running at well over 300mm per second (which was crippled by the extruder). My thoughts on printers have changed a bit after having interacted with both of them. Want good, consistent prints, buy/build a stiff, solid printer, something almost every other printer lacks. The Griffin and Ultimakers are the only ones I have seen that could really handle being moved and not require a lot of maintenance (I have seen both turned upside down while printing).
Printers I have experience with:
Ultimaker, looks older and sort of like a homemade kit, but in the right hands, it's an amazing printer.
Ultimaker 2, very nice looking, and the impression I got was that in the right hands, would be a really nice printer. I helped with some early tuning on one.
Griffin delta, personally, best delta I have seen. It's fast, as big as you want, solid and cheap too. Documentation is coming along, auto-calibration is in the works. Kits are rumored to be in development.
Lulzbot Taz, it's a nice printer, but the nozzle is fickle and the bed is easily knocked out of alignment if moved.
Rostock, they work, but lose calibration pretty easy. Switch to rod ends and make an actual frame. (May as well just make a 3DR)
Kossel Mini, nice, but entirely too flimsy, and not the cheapest delta. Without auto level it would be as troublesome as the Rostock. Maybe more.
Rostock Max, change the diagonals, it's a decent machine. I haven't seen how the new ones are, but the old ones were a nightmare. Not one of my favorite machines, if only because of the excessive use of wood and junk diagonals.
PrintRbot original, it's an old design, and we have come a long way.
PrintRbot Simple, keeping it's cost in mind, it's an amazing deal. It's high maintenance though and over time the accuracy will drop off. Would I buy one? yes, then make a different printer and cannibalize it for the parts. LOL
Prusa, just no. Like the Printerbot original, it's an old design that requires constant tuning. The I3 is supposed to be miles better, but I just can't get excited by it. Probably because it's not something the average person can build without buying the main assembly (it uses laser or waterjet cut aluminum plate), while other designs can be done with common tools. This is a big departure from Reprap ideology.
I have yet to see a Makerbot in person, which is about all I will say about them.