I've read threads about the SOC crashing, but this is pretty ridiculous... I've seen it do this at various SOCs between 15% and 30%, it's not great. I have an AIO, firmware 0.605, calibration finished properly after install (maybe it needs to happen again?).
Welcome to the p!ssed off AIO owners club. Have a look at jellybabies battery calibration thread.
You are on quite an old firmware. I'm on .608 but it also has it's problems namely charges to 100% with less that what was discharged (also from 100%), which leads to one confused battery that unexpectedly runs out of energy
That last 30% hasn't gone anywhere. It is either still in the battery or was never in there in first place. The software has an incorrect view of what the state of charge is. Either it though it was full when it wasn't or now thinks it is empty when it isn't. You can't lose 30% of the energy instantaneously without an explosion. Losing it over hours would even release a lot of heat.
Had the same AIO unexpected SOC fart the other night. I’ve been running PredBat very successfully for about 6 weeks, I ditched it fully a few days ago as I’ve moved to a different minimal export strategy so just need the GE app. Might just be a coincidence that as soon as GE got back control of my AIO I see the same % drop. I realise it can’t just discharge that much energy into thin air so like you say it was either never there in the first place or it’s still there.
I’m still on inverter firmware 607. Battery firmware 6. I’m not updating until they make a significant firmware improvement, that is assuming I have any real control over my own AIO!?
You would think one of the basic 'no brainer' functions of a battery management system is knowing the actual state of charge of the battery.
Yes but it isn't easy looking from the outside of the battery what its state is. There is only voltage and temperature that you can measure. There isn't a direct measure of energy stored in it.
Done it again today, all battery strategy goes out the window when you don’t know the true SOC. 
Bullet dodged…
You are on quite an old firmware.
Interesting, thanks; this was installed on 4th Oct and it did a firmware update as part of commissioning. How do I trigger an update? There's nothing obvious under the "software" tab in the GivEnergy portal, although it says "D0.605-A0.605" for "Inverter Firmware Version" (and "battery firmware version" is "7").
Honestly if what im seeing from 607 firmware is right and its moving the ceiling down so you never ever get a full charge and keeps lowering the max capacity of the battery you are better off staying where you are and doing the math on charge/discharge to work out when its going to go kaput. At least on older firmware when I did plan for it I got the full capacity of the battery back whilst it started ticking down.
Im expecting to have no battery to charge within a week and contemplating powering it down via the gateway breaker for when 4pm starts im going to have that little capacity left.
Yowzers
Reading the other threads again made me check the SoC drops against the battery voltage, and it seems to happen around 285v, so I guess that's a better predictor of SoC than the percentage readout?
Yes but I can predict that based on charge/discharge kwh put into the battery, due to the portal using 1 decimal place I have only been 0.2-0.5kwh off on my guess when the battery would tank.