ClauBebe

Unfortunately, it's not something I can predict in advance. Since these patches don't truly cure the problem (only Voltronic with the source code and extensive lab equipment can do that), it's a matter of trial and error.
It sets the minimum value for a firmware variable that represents the set point for the MPPT voltage. In this context, set point means the point that the control system is aiming for. So if that value is 150.0, then the MPPT won't ever be aiming for say 140.2 V, even if that is the true Maximum Power Point (MPP) for the present conditions.
Starting is sometimes different to running, but with Voltronic's firmware, they don't seem to do anything special for starting, apart from forcing the MPPT voltage set point low under some conditions, and ironically, this seems to be the crux of the problem. Too often in cloudy conditions, it's acting as though the sun is extremely weak, and it has to drive the panel voltage down to extract enough power from that very weak sun to get everything started. It has no memory of operating at 90% of rated power two minutes ago, for example.
These are "families" of patches, all identical except for the minimum MPPT voltage as mentioned above. With my latest patching "technology" (developed in conjunction with colleague Weber who helped tremendously on the fully patched Axpert MKS and Axpert King firmwares), it's relatively easy to produce these sets of patches. From the way that users have responded, it looks like this is the way I should do all future MPPT patches, at least until some breakthrough comes along that allows for a "one size fits all" patch. As it happens, he has no interest in these MPPT patches, so it's just me for these. And I don't own an inverter with the problem! None of these MPPT patches would have happened without the help of owner @Georg594 (thanks, Georg!)  Of course, he only owns one model, but at least that one has the actual problem. He "translated" his patches to one other model, from memory; I've been extending this to several models' firmwares, because, well, that's what I do 🤓. There aren't many of us that can do this sort of thing, so when I find the time, it seems only right to use this uncommon "superpower" for the common good.
Time has been a bit tight for me just lately; we're having a heat wave here (37-38°C and ~80% humidity, interspersed with thunderstorms) and our air conditioner has packed it in. Before anyone makes a sneaky comment, it's nothing to do with firmware; the motor bearings wore out after 19 years of operation. So it's been a bit of a priority to organise a replacement.
 
Thanks for the suggestion/sentiment, but of course it will always be completely optional for people to support my efforts.

I also managed to install the new firmware 56.07
  and I chose the 150V version (Reflash dsp_x56.07_patched - 556.07_150V) to see how it behaves compared to the previous Coulomb patched version 56.05 (OG-10 DSP Reflash_656.05 patched) which I have had for more than a year and which perform extraordinarily well compared to the mess of the original firmware that was stuck in 90V forever on a cloudy day !
I will come back with an update after a few days of using the new  firmware!
Thank you Coulomb and Volker 
 
 

I can't wait to test it, TX for the hard work!!
My setup is 2 string on 375w * 10 pic's Canadian Solar and 25Kw li ion battery Nissan Leaf with Orion Jr 2 BMS and Victron Smart Shut for monitor the pack in solar assistant 

Just went through this amazing thread post by post. I just have to say that I am amazed at the wealth of information and learning curve on my end. Thank you for all the contributions. 
@Coulomb can you point us to that "thread" with a link, i was not able to find it.

Yes, I intend to patch firmware version 57.60 with the equivalent of patched firmware version 656.05 for the Max II 10 kW models.
57.60 seems a long way from 57.01; is yours a relatively old model? What does your manufacturing month small sticker say?

With QFW 656.05 I caught a cloudy and rainy day with very weak sunlight
Compared to the Solax that I monitor and which works at maximum capacity, which today produced only 0.80kWh all day, my system produced 0.90kWh, something that never happened to me with the original firmware, which remained stuck in 90v for a day like this, and it didn't produce anything only if I keep resetting it manually
I am attaching the comparative graphs between the 3 QFW 456.05, 556.05, 656.05


and here the whole chart for today, cloudy and rainy day with QFW 656.05

 
I'm curious to see Volker's graphs, which can track in real time the behavior of the 2 firmwares QFW456.05 and QFW656.05

Here is patched firmware version 556.05 for the Axpert Max II 10kW with the older '2809 DSP processor.
It implements:
* Fixes the premature float bug.
* Displays as version 556.05; I think I messed up the version number for what was supposed to be 456.05.
* Implements Georg's current threshold patch, 0.50 A -> 0.05 A.
* No longer stops setting the direction flag when the current is low; I decided it would make no difference.
* Every 5 minutes now (was 1 minute), it checks each MPPT voltage. If below 175.0 V, the MPPT (1 and 2) is "reset" by changing the direction flag to 1 (was to 0 in 456.05) and sets the MPPT voltage threshold to 200.0 V. Setting the flag to 1 should hopefuly fix the "glitchiness" of the PV voltage.
* MPPT1 and MPPT2 are handled in separate patches now, in case they get out of sync, which could have been another source of glitchiness.
* Fixed the bug where I was testing the solar voltage reference against 175.0V, instead of the latest actual voltage measurement.
Use at your own risk.
OG-10 DSP Reflash_556.05 patched.zip

Man, i love your install, that extra lcd display on the Max is beauty😍

I can't wait to test it, TX for the hard work!!
My setup is 2 string on 375w * 10 pic's Canadian Solar and 25Kw li ion battery Nissan Leaf with Orion Jr 2 BMS and Victron Smart Shut for monitor the pack in solar assistant 

I think it has. I should be doing the resets after an adjustment cycle, when the accumulated power is set to zero, so the next cycle, you'll get a clean measurement of the difference between 175 V and 200 V. As it is, it's basing its direction on the last measurement when it was bumbling down at below 175 V, and the measured power was likely decreasing. This change is also pretty easy, and should result in better smoothness. I should split the patch into two separate pieces as well, one for MPPT1 and one for MPPT2, since in theory they might get out of sync with their counters.
This next patched firmware should be a good improvement. A good thing too, as I'm running out of patch version numbers: 65 605 representing version 656.05 would be > 65 535, so that won't fit into an unsigned 16 bit variable. So 556.05 has to be the last patched version. No more bugs! 🙂. Of course, I can resort to a different numbering scheme.
[ Edit: Sigh. Had wrong firmware version number to patch. ]

I've just realised another possible reason for the jitter. In the previous patched version, when the MPPT reference voltage was set to max (450V), the direction was set to 0, meaning that we expect the PV power to increase as the PV voltage is reduced (we are above the MPPT voltage). This makes sense.
But in this patched version, when we set the MPPT to a middling voltage (200 V), we do this because the PV voltage is low (below 175.0 V), so the direction should really be set to 1,meaning that we expect the PV power to increase with increasing voltage (we are still below the MPPT voltage). This might occasionally not be correct, in which case we'll get the occasional glitch, can't be helped, but in the common case where the MPPT voltage will be over 200 V, the assumption will be correct, so the PV voltage won't go in the wrong direction most of the time. Fortunately, this is a trivial change.
I came across this while commenting the many many firmwares that I have now, attempting to check for a difference between MPPT1 and MPPT2. I strongly suspect that there is none, and the differences in behaviour are due to other factors, but I may as well finish that commenting and comparing, just to be sure. Who knows, it might even shake out more bugs, in either the patched code or the original factory firmware.

Yes, you have 56.xx, so you can use the patched firmware with versions y56.05, where y is 1-4 so far.
But note that I've found a bug in 456.05, and we're still comparing Georg's and my approaches, so it's probably worth waiting for the dust to settle.