It seems though that enemy AI prioritize threats incorrectly, or rather, not smartly. I am just guessing here, I haven't checked any code what so ever as I dont understand it, but I am just guessing. AI prioritize their targets based on who is the greater threat, and that could be determined by few and simple variables such as type of weapon/role and line of sight perhaps?
Of course targets should be prioritized by threat, but the distance as well, and not to mention direction, and who is engaging that AI unit specifically.
For example, a support gunner will be a bigger threat than infantry man, naturally. But also an infantry man 100m away is a greater threat than a support gunner 150m away. AI should react and co-ordinate like such. In the example if one (1) AI unit was engaged by both players coming from the same general direction, it would first seek cover, and then had to decide who to engage, then it would choose the infantry man over the support gunner. However if gunner was due west 150m and infantry man was due south 100m, nearby cover would not help much, so it would fall back and seek cover far enough until both units are within general direction, and then engage the bigger threat, whichever is the one closest.
In this case, AI wont act silly and just ignore the player to prioritize and engage what it is programmed to believe is a greater threat, it should be aware of the presence of a unit that engaged it for as long as it hasn't confirmed a kill or can see the unit. Anyone else that engages said AI, will force AI to seek cover or fall back, and then reasses and prioritize.