You can just slap a dial gauge on the gantry and move the X/Y manually to see exactly what the deviation is. A decent second-hand dial gauge on eBay will run you $20 shipped.
If you get into the weeds, there is not an accurate method of triggering any form of mechanical stop that involves touch or a hall effect probe. You must get into optics for real accuracy, but that is nonsense for the materials and scope of printing. You would need to eliminate many other variables like the filament accuracy and how backlash and step accuracy are eliminated as issues.
As a former owner of an auto body shop with employees, most people do not know what clean is or how tooth is required. Like isopropyl alcohol has its place but is ultimately extremely weak at real cleaning problems. In automotive paint, silicone is a major problem. It primarily comes from tire dressing that makes them look slick black. The amount of effort it takes to remove that junk for automotive quality work is insane. Most chemicals just push the junk around but leaves or dilutes the issue often making it worse. One of the big tricks in automotive stuff is (to use a chemical cleaning step first but -) a few drops of dish soap in the wet sanding bucket. The light soap will keep the sand paper clean and working longer, but makes most work also cleaning work. Anyways, dish soap can be very effective. Acetone occasionally on a surface is also effective. Virgin lacquer thinner is the strongest common solvent but it can react with lots of stuff and you are unlikely to find true virgin solvent. The recycled stuff has a paint stripper component in it that will cause epic nightmares and reacts with almost all plastics. Acetone is much cleaner and consistent unless it is sold for junk like nail polish.
The general rule of thumb is to assume a mechanical tooth adhesion is the primary form of bonding unless there is a catalyst involved (2k urethane/epoxy primers/clear). That rule can easily apply to 3d printing and bed adhesion. I see a lot of the same types of effects from different surfaces and filaments. In automotive paint, there are even special adhesion promoters like Bulldog for spraying plastic parts ahead of other finishes. I had other adhesion promotion tricks too, like a mist coating of clear coat. The main trick with all automotive paint adhesion is to know what grit or “tooth” each thing you’re spraying wants to grab onto and prep accordingly. So in 3d printing I use a similar approach with the general safe bet of sanding my smooth build plates to 600 grit. With sanding, do not start dirty, like you’re trying to embed junk into the surface. Start clean, then knock off the shine to a smooth and consistent matte finish on the entire surface. When it comes to sanding like this, edges and any anomalies are absolutely forbidden to sand. Never touch your edges until last when everything else is done. Edges are always thinnest and most vulnerable to causing issues especially for the inexperienced. You match them to the rest of the matte surface carefully at the end.
Clean a smooth build plate with acetone like once or twice a year and then sand it to matte, clean that with dish soap, then alcohol with each print. That will completely eliminate contamination as a cause. If you have old skool clean glass with no coatings as a build plate, sanding is optional because you can use something like lacquer thinner or less effective acetone to get it absolutely clean.
Perfect first layers are possible with enough fussing with the software. If you really want to level the bed with hardware, use a dial gauge clamped to the extruder. That will remove all of the averaging and inaccuracies from probing if it is a quality gauge that is smooth and not sticky. You would need to get into optics for true accuracy like with closed loop control systems that are an order of magnitude more expensive than 3d printers. 3d printers are precision machines with no accuracy. The 0,0 home location is always slightly different, but all measurements are based upon this location. This issue becomes relevant with IDEX and CNC. Going well beyond these – in optics accuracy requires a defraction grading and alignment of light wave patterns. I so want to get into that one to grind my own telescope mirrors. Typically accurate machines use a flag of metal sticking out somewhere at a known location and an optical encoder switch that gets interrupted without anything touching as this is typically the closest you’ll get to real accuracy down to the clock and instructions timing of the interrupt routine in the microcontroller.
If you have v-roller wheels on extrusions, one other major potential issue is that extrusions have a relatively large twist tolerance component in their specification. It is extremely difficult to detect this kind of twist, but it is a major potential issue. It generally requires a high metrology grade granite surface block and parallel sticks to measure twist in a precision instrument’s linear bearings… as far as I understand it. I have seen such things being measured but have never done so myself.
“Randomize seams” uhh no, design parts with seams in mind if they are critical. I often add a small 0.3mm double chamfer “zipper” on a surface because it will accommodate the inconsistency and reduce it by forcing the root inside the body. I’ve made my own infill structures and patterns lately too.