Lindsay

Yes, I have the Mecer 3KVA 24V Plus and it's setting 04 in the menu, SdS: Savings mode disable (default) and SEN: Saving mode enabled. In savings mode it pulses about once a second and tries to sense a load, and as @Flouw says, the load has to be over a threshold for the inverter to turn on, about 100W in my experience.

Yeah, even at a couple amps they seem to get very hot, so will attach a heatsink, or will point a fan at it  

Thanks everyone for the responses, I didn't expect so much feedback!
While I was going to try the light bulb/resistor option, I now realise that I do have some buck converters that do CC/CV which I should be able to use. I bought these when my solar was DC only and was using these to convert the 24V (from deep-cycle at the time) down to 12V to use with lighting, router etc:
CMU DC/DC BUCK/LIPO CHG 1.25-36V - Communica [Part No: CMU DC/DC BUCK/LIPO CHG 1.25-36V]
So maybe it would be better for me to use this on the low cell? I could then set it to 3.6V and the current to 3-5A. Has anyone set current on these before? The website says I should set my multimeter to current and put it in series across the output...would that not just be a short circuit and damage the multimeter or the buck or would the current protection on the buck then kick in? (I think I've have seen people measure the current output of solar panels in this way - but I always worried that something would get damaged).
10A current with a multimeter measuring the output short circuit current file (directly to the two leads to the output can), while adjusting the constant current potentiometer allows the output current reaches a predetermined charge current value;  

Only getting time to look at this now, thanks for sharing
 

The guy who runs the powerforum store told me (at least for the 24V version) that it had been withdrawn from supply due to quality issues and when later I thought he had it for sale again he called them "untested". He also told me over the phone that they were 2nd life batteries. 
So based on that and what you say, it doesn't sound like a good option
 
I'm still trying to find reasonably priced LifePO4 options for a 24V setup. I was about to buy from lithiumbatterysa, but with insurance and shipping to Cape Town it adds R 2000 to the price, which puts it over my budget. Would much rather buy something already in Cape Town, but struggling to find anything reasonable 
 

I just checked and last night it was 47Ah, that's across all 4 strings, so about 12Ah per string. I have a fridge, freezer, router, some security cameras, LED lights and chargers for phones and laptop that run off it, so not a heavy load. I have a slow cooker, induction stove and mini-oven, but only use those in the day when there's plenty of sun. I should probably look back properly through my logs and see how it ran last year this time to do a proper comparison.
I do notice that sometimes the battery voltage reported by the inverter goes a bit over 30V (pretty sure it has done this since I got the inverter). I think this is brief and just when it's switching modes (absorb -> float) etc so I haven't worried about it.
I gather when I find the battery bank is really struggling I should switch out the whole battery bank in one go. I seem to remember the manual saying you shouldn't use any battery type other than lead-acid. However I gather if I get a bms and adjust the absorb and float voltages would it be ok to use a different battery type (lithium etc)?
Thanks for all the help, I'll let you know what I find!

Thank-you Lindsay.
I have written my own crc function to take into account the Axpert's idiosyncrasies and have fewer dependencies. After spending some time fiddling with Node Red  and node.js I have developed an aversion of libraries and dependencies unless they are well written and documented. It appears Voltronic use a chipset that masquerades as a Cypress Semiconductor chip. From a little saunter on the web Cypress seem to disown VID 0665 PID 5161 which leaves most of us fumbling in the dark. The Python 2/3 changes how strings are handled does not help either. I prefer how Python3 handles strings with text and bytes being discrete but does give one a headache when you trying to get an  officially undocumented chipset to work.
Thank-you Lindsay for rolling several large boulders out my way. In the mean time I have dusted off my 30 year old programming skills. I hear a cacophony of laughter coming down the ether. I cut my teeth on Basic, Cobol and eventually a little bit of Pascal and DOS doesn't stand for Dark Outer Space.😆  
Since I am so rusty I usually write 2 lines of comments for every line of code. I cannot abide not understanding code I have written/copied and end up dissecting github code. This I suppose defeats the object of repositories but too many folk (myself included)  borrow code from a different source with scant  regard to its robustness. This is in no way a slight on Lindsay's contribution for which I am most grateful.

Ok, so I seem to have it working now in Python 3.
After Gnome mentioned the encoding and I skimmed the articles above I looked at the encoding and changed my original code, which was based on nikitto's post from June 4, 2017 in same topic:
### code above removed, see nikitto's post from Posted June 4, 2017 in same topic ### calc_crc(comando) print('Command='), comando print('CRC='),crc crc1=crc[0:4] crc2=crc[0:2]+crc[4:6] crc1=int(crc1, base=16) crc2=int(crc2, base=16) scomm = comando+chr(crc1)+chr(crc2)+'\r' bcomm = sCommand.encode('ISO-8859-1') fd = open('/dev/hidraw0', 'rb+') fd.write(bcomm) r = fd.read(nbytes) j = r.decode('ISO-8859-1') s = j.split("\\") print s i = s[0][1:].split(" ") print i fd.close() So I open the connection to read/write bytes (not strings) and encode the string before passing it to the write and decode when reading (and removed the .encode('string-escape') from the original code).
This seems much more reliable and I didn't see any errors when running this code. I hope this helps others that are trying to get it working in Python 3.
Thanks again Gnome.