There are 2 reasons for having 2 instances 
- Migrating from one system to another without downtime
- Fail-over configuration
I think I may have found a solution for this particular case: as mentioned in Loop when using a knx2 switch-control channel - #10 by miro-s, setting
Switch wallswitch1_long { channel="knx:device:bridge:generic:wallswitch1_long", autoupdate="false" }
actually causes the previous OH1 behaviour: OH2 receives the command but doesn’t answer.
Now I think I am even more confused about when to use switch vs. switch-control. If I want to e.g. have a dimmer that has a real KNX actuator and a real KNX wallswitch controlling it, and I want to distinguish between the actuator being set from OH2 (through the UI or a rule with postUpdate) and a change triggered from the KNX bus (e.g. the physical switch), do I need to duplicate all my GAs?