Sorry, i don’t know how your comment should help…
As mentioned above, the problem is that as soon as a grid code is set, the inverter does not work anymore with a AVR generator even if the input is configured as generator.
And we do not need AC out 2…
I always have a 1.5kw heater (resistive) connected to my conventional generator, without this my multiplus 12/3000 will not connect.
this sounds like otherwise the input frequency is too high…
No, its probably the waveform.
But as i have a solution i don’t really need an explanation to be satisfied, I’m not that nosy 
Also think it could be related to the waveform. That with the heater as a baseload is the same thing, but likely more crude than the thing with the waveform-sensitive ups and the freewheeling compensated induction-motor i stated.
Question - did you already had a look on the waveform generated of the AVR Generator with an oscilloscope?
And did you try it with disabled UPS mode? As i stated - disabling it does not remove the function of a battery backup if power fails - it only prolongs switchover time a little bit - but with the benefit of being less picky regarding the input waveform.
I´m with you that the charger-only variant is no real solution 
I did not but AVR generators usually output a perfect sine wave by design (at least if we are talking about high quality generators)
I did, no changes
I am still convinced that the problem is that while AVR gensets output a perfect sine wave (the grid consists mostly of AVR generators) they do not output a perfect frequency and there the inverter makes a mess…
I will now try to get a official answer from victron and see what they say
This is so far the official answer from a Victron dealer:
If a generator is used as a power supply, a MultiPlus will work without any problems, then of course do not use a grid code as it is not connected to the grid.
*If there is a grid, then use a Quattro with an extra input for an backup generator. *
Grid and generator on one AC IN do not work.
In this case, at least the generator would have to supply a perfect current equal to the grid and this constantly, then it would perhaps be possible.
Wich i replied with:
Thank you very much
I have made the following test:
(with both a 100 kVA and a 20 kVA generator)
Grid code Germany, AC in 1 LOM set to generator, generator connected to AC in 1.
But it did not work!
And then added this information:
I did the test with 3x quattro 48v 8kva (1 per phase).
Then I did the test with 6 5000 multis (2 per phase).
I only tested the multis with a 20kVA AVR generator, the generator oscillated extremely (0-100%)
I tested the quattros with 100kVA and 20kVA, the load was extremely asymmetrical and not constant.
Currently waiting for a answer…
i wanted to do a test with the multis…
How should i setup my system so it only charges the batteries from MPPT or power from AC out?
As far as i can tell, the culprit is “…or power from AC out?”. Feeding/Charging on the AC out is an ESS thing, specifically the function of supporting AC coupled PV on the AC out. And for ESS a grid code mandatory. And as soon a grid code is in place, you would probably expericence your problems with the generator(s) again. 
But nonetheless i would have a look at it with a scope or power quality anazlyzer, especially at the region of the zero-corssings.
Because of a theory/possibility:
While doing their ESS thing the Multis/Quattros have to keep in sync with the power feed somehow. Althought i dont´t exactly know how victron does it, the usual was is by zero-cross detection, usually by an op-amp based contraption often in combination with some hardware or/and software based refinement. At least so i did it for the thyristor rectifier synchronisation in my inverter…
Imagine the following: If you have glitches, spikes, small harmonics or whatever in the zero-crossing region it could cause early or late firing of the zero-cross-detection and thereby causes the system to mis-sync. Even a one degree phase angle mismatch could have huge implicatons in the resulting current draw and thereby loading condition. And it could cause the described unpredictable non-symmectric oscillating power draw.
Although i think you already understand what i mean, here a (perhaps slightly extravageted) picture.
Without ESS/Grid-Code there is no need for the whole grid-parallel and synchronizing part - the Multis/Quattros simply take what they get and use it. Therefor this couldnt happen in the first place and could be perhaps a reason why it works without problems without a grid-code in place.
At least with the Quattros, feedback from AC out works even if no ESS is configured.
I will get a oscilloscope and continue my research.
I just also had another idea.
With the Multis i always had the grid present but set them in inverter mode so it disconnected from the grid.
But i have a theory that they probably stay in sync with the grid all the time even if disconnected. This would obviously result in frequency shifting what the generator cannot handle…
Will give feedback once i have a oscilloscope in my hand.
Didn´t expect that feeding from AC out without ESS works. My bad - and good to know! 
Your theory about staying in sync with the grid als long as it is present on the input even when disconnected absolutely makes sense. With this, the system could switch near instantly to grid if necessary instead of needing to go to the sync process. Could also be verified with a scope and two siutable differential probes.
today i did tests with a osciloscope.
first of all. if you just put the inverter in Inverter only mode, it stays always in sync with AC in.
the only way to get it out of sync is by physically cutting the power on AC in, then it goes to solid 230v and 50Hz
already being happy i did another test with the generator on AC-out… without luck. the genset synchronises, closes its circuit breaker and as soon as this happens, the AC out is no longer a sine wave. it is more a Erinaceidae…
meaning the 50Hz remain solid but the voltage cuts at about 120-180v. causing huuuge currents on the generator. The inverters turn off after a very short instant because of overload…
so what we know now:
ESS/Gridcode and generator, no matter if AC in or AC out does not work.
no ESS/Gridcode and generator, no matter if AC in or AC out works
As long as we are Grid Parallel it does not matter if the Generator is on AC in or AC out and it does not matter if ESS/Gridcode is enabled or not.
Interesting.
Confirms the theory that the inverters stay in sync with the grid as long as it´s connected to the input, even in inverter-only mode. Did test this out of couriosity yesterday on my system – same result.
I must say i have a bad feeling about feeding via AC out with ESS/grid-code. Think it wasn´t ever made for genset feeding where the inverters are the grid forming part. Probably works fine for AC-PV feed because that can adjust freq. and phase near-instantly and a genset cant. I think it works without grid code / ESS because in essence you take the AC in relay out of the equation. It does not have to sync, it simply takes what it gets.
Question – did you scope the test setup with the genset on AC-in with the genset as grid-forming unit and the inverters with grid-code “Generator/Other” and ESS (but without LOM!) and UPS mode disabled? And only the generator – no grid physically connected on the other AC-in.
It would be interesting what the waveform of the genset looks before and after sync when the system tries charging in this config. Also pictures of the waveforms would be nice.
FYI – my setup for feed switchover is as follows: I have a contactor before each AC-in. They are interlocked so only one can be closed at the same time. I know the Quattros should do that internally but i´m a friend of redundand safety and even more safe interlocks. So if i switch over to generator on AC-in-1, the grid on AC-in-2 gets disconnected and vice-versa.
Additional info although it´s unlikely that it has somthing to do with it - but this are my settings for reactive power regulation on my quattros. (cause the grid company wanted it that way…) It is also active for my generator input because i have the VDE grid standard on both inputs, only without LOM on AC1 for the generator:
“Standard” would be cos. phi 1. As far as i know it´s for stabilization purposes that the inverters behave a bit inductive or “underexcited” if the voltage is high and a bit capacitive or “overexcited” if the voltage is high.
