Making computers do things, physical things, in addition to crunching numbers in mission critical process control applications is what I get paid to do. If something does go wrong, either from an operator error or a bug in my software, I have to make sure I completely understand why so it cannot happen again. For this reason, I have a handle on this. The ECUs used are extensively tested, and are not exclusive to one mfg. I had a Cannondale that I did a lot of my own custom mapping on, as well as a lot of electrical hardware re-engineering. That bike used the same ECU (Sagem, with an Optimum operating system) that was used in some Renault cars and the early Triumph Speed Triple. The bike got a bad rap for EFI reliability. I identified the problems fixed them, and it was solid as a rock thereafter. Some of this was electrical, some was thermal. Its all in the implementation and packaging, it has to be thought out very well and then life/torture tested. The Cannondale used off the shelf components like the fuel pump, and an external low pressure fuel filter that was easily inspected and changed. Never any injector clogging problems like the KTMs and Bergs. Due to the reverse motor, injectors right behind the radiator, airbox with temp sensor above the head, and header under the tank heating the fuel, there were some challenging thermal fuel and sensor heating problems to solve when the bike was ridden in a tight woods environment. Open air fast MX no problems. From what I see of the Bergs after a good look, with the high header and tight packaging, it doesn't surprise me that mystery problems pop up occasionally.
I think that 2stroke DI, while perhaps more complicated than EFI, would have an advantage over 4stroke EFI in that there is much less heat to deal with. Dirt is the easy part to deal with, heat is the devil.