Here is the problems as best as I understand it: the difference is that it is more difficult to regulate power accurately when the flywheel spins more quickly. Let me assume that your trainer uses electromagnets to generate resistance, and this is done by moving magnets closer to or farther away from the flywheel. Braking force is usually applied in discrete (digital) steps, so you only have a limited amount of accuracy. What is regulated by stepper motors is the distance between the flywheel and external magnets.
At higher flywheel speeds you need smaller steps to achieve the same level of accuracy in the applied braking force to produce the same resistance. That is because the flywheel energy to leading order is quadratic in the flywheel speed. At lower flywheel speeds, you inherently have more accuracy. Think of it this way: at high flywheel speeds, the trainer has to pick between 290 W, 300 W and 310 W as discrete power levels, whereas at lower flywheel speeds, it can do 290 W, 292 W, 294 W, etc. This is a bit of an exaggeration, of course, but I think it accurately brings the point across.
The second difficulty is how often you measure power. On-bike power meters give you one measurement per second. Some power meters (e. g. by Rotor or SRM) have special modes that allow for more frequent power measurements. Rotor uses it to determine your pedaling smoothness and to recommend settings for its oval chain rings. I don’t know how often the built-in power meters of trainers measure power, though. But the lag is an inherent problem. Humans don’t apply power very evenly, so you need to smoothen the power data. This is why for most intents and purposes, having 1 measurement per second is sufficient. Smoothing introduces lag, which makes applying resistance difficult. This is an old engineering problem. If you just go by current data plus the trend (i. e. the slope/derivative), you can easily induce oscillations. So you need to average (integrate) and add some more smarts. This works way worse if you have less discrete levels of braking that you can apply. Roughly speaking, these are the two main issues as I understand it. (I’m a physicist, not an engineer, so if anyone more knowledgable can chime in, please do.)
That is why TrainerRoad recommends you use your small chain ring for Erg mode. However, there are cases when you may want to use the big chain ring nevertheless, and that has to do with trainer feel. A faster flywheel will feel more like riding fast on the road, a slower spinning flywheel will replicate the feeling of climbing. So with a slower flywheel, if you let off the power, the flywheel will spin down more quickly, which is similar to you slowing down very quickly when you stop pedaling on a climb. On the flats at high speed, you can coast for quite a bit longer without significantly slowing down.
PS Let me add that trainers should get better and should be able to handle you being in the big ring better. On the other hand, all technology comes with inherent limitations, and once you know them, it is usually smarter to work within them.