First one is to simplify the setup as much as possible and get it to a point where it is good enough so the latency is not very noticeable. I have found this to be the best solution for many scenarios, especially if you play instruments live during the performance.
Solutions involving sending different timed clocks can often introduce other issues depending on the solution and the routing.
This could be a solution but if you send audio too much back and forth I guess it would still be difficult to get consistent results.
If I understand you correctly, the ES-9 is used as the only audio interface? If so, I would start with limiting the use of its inputs and outputs so that the inputs are only for sending sound to bitwig and use no outputs other than the final output. This would mean you would have to go analog out from modular to OT and give up on sending vocals (or whatever sound) from bitwig to OT.
I understand that you have something particular in mind so that changing things can change your idea, but it can be a good idea to try to make the most essential parts of the system work first and then address the problems one by one. That way you won’t end up getting a multi clock solution only to find that there are still issues caused by the routing.
Hope you manage to find a solution in time
(I apologise if I am not explaining everything clearly as English is not my first language)
That’s a really kind proposal unfortunately I’m in Sweden but I still wanted to thank you for the offer, going to consider the CLOCKstep:MULTI when I see it in stock again
Updates: I tried some options you all proposed here, including routing all audio through the DAW.
Sometimes I got Bitwig to compensate the latency pretty well even without routing all the audio through but it’s spotty, I will see the latency compensation notice being displayed in some sessions (about 55ms), other sessions don’t show any indication it’s being compensated and I definitely hear mistimed sounds with a simple test keeping the modular kick on a 4/4 and open hats from the AR on the off beats.
Routing all the audio through the DAW does help a lot as expected, it introduces the same latency overall and since I’m not playing any acoustic instruments live it would be totally okay.
But it’s so finicky that I have given up on setting it up this way for the upcoming set, it’s taking my confidence away and making me question a lot if everything is on the beat as it’s supposed to when practicing, also questioning if everything will work as intended on the day so in the end I’ve decided to ditch the AR for this one and have setup VSTs to cover the percussion tracks I wanted to use the AR for, and going to use the Oxi One to sequence the VSTs.
Had a short 1 hour session and everything sounded in place.
I will not give up on making this work for the future though, just going to stop wasting time with the technical issues for the moment because it started to stress me out
Thanks a lot for all the help, for future people who finds this thread: the solution is very likely there in one of the suggestions, I have just ran out of time to properly analyse each one more deeply to find one that could be trusted.
yeah sometimes it’s the best to downgrade when not enough time.
Live Performance is sufficiently stressful to not let the latency and setup issue adds on top right decision you made here
There is one other thing you can try, and I’m just spitballing. I’ve done this before, but not for this reason.
Use an external clock to generate both MIDI Clock and Analog Sync. Have the DAW follow the Analog Sync on its Audio Input. You can do this even without DC Coupled Audio inputs on your audio interface. It takes a plugin in Ableton that is included in Max Live; I don’t know if there is a Bitwig equivalent.
The theory here is, with the external gear following the external clock and sending all of their audio into Bitwig for processing/monitoring, AND Bitwig following the same external clock on an audio input, then the external clock and the external audio will have the same input latency in Bitwig.
I know you are probably tired of hearing about CLOCKstep in this thread, but this video is honestly the best way I know to demonstrate what I’m saying. You can get other gear to do this besides CLOCKstep, all it needs to do is send both MIDI Clock and Analog Sync.
unfortunately I’m in Sweden but I still wanted to thank you for the offer, going to consider the CLOCKstep:MULTI when I see it in stock again 
and have setup VSTs to cover the percussion tracks I wanted to use the AR for, and going to use the Oxi One to sequence the VSTs.
right decision you made here 