Steinar, I've been monitoring ROS on-the-air and I've done some testing with the HF path simulator. In my opinion, it's about as good as one would expect from an MFSK mode with a relatively slow baud rate. Tests show that it will outperform RTTY and PSK31 in poor channel conditions (most MFSK modes do) but it does not appear to be as robust as Olivia.
For example, it is less tolerant to Doppler spreading than Olivia so it's less likely to do well when the ionosphere disturbed. This is especially true for polar paths and the low-latitude ionosphere where Doppler spread is more of an issue. While the mode performs well over HF, the additional bandwidth doesn't appear to have any throughput advantage over other modes that use less spectrum. In fact, path simulations indicate that there is no difference in throughput between ROS 500/16 and ROS 2250/16. Tony -K2MO