It shouldn’t be. There is no way the device can know… 
And I’m also just speculating. The most basic at-face-value reading of the situation, is that the device is doing it’s own earth fault test to make sure it is safe to reset the device, and this test is failing. This suggests that even after solving the bonding, it might still not fix the problem. But there is a small chance that it is actually testing the bonding… I don’t know.
No problem Niel, thats just the impression I got from our conversation. Apologies if I was wrong.
It’s still sad that I could not get to your place, as I am sure we would have found a solution.
Just to highlight how a RCD operates: It doesn’t connect to earth at all.
It operates by checking the current that flows between live and neutral (in a single phase system) is balanced.
Typically leakage current goes to earth but it could be anything that feeds in or drains current to your system.
I’ve seen older 3-phase earth leakage switches that do, and I still have such a thing in a box somewhere. But all the modern ones are merely residual current devices with no earth connection. If there is more than 20mA (or so) difference between the two lines… click!
That’s why a poor earth connection is more likely to prevent it from tripping than the other way round. If your earth connection is poor, that means that when there is an earth fault (eg a human being with his bare feet standing on cold concrete touching a live wire), then the additional impedance caused by the poor earth reduces the leakage current, which makes it less likely that the RCD will intervene.
My understanding (which is a layman’s, very limited, and not to be quoted, understanding) is that you get ELCBs which you would use in an installation with a high earth impedance that would trip on a potential difference to earth, not current, while the RCDs by definition would be tripping on a current leakage.
While the above would seem like a statement, it is really a question and I’m happy if someone educates me on the topic.
Ok, that makes sense, but by bridging the earth (input and output) at the inverter, am I not overriding the inverter’s connection?
Ok, things got strange. I tested the voltage between earth and neutral on the output of the inverter and it reads zero (understandable as it is pretty close to where they are bonded). At the multiplug in the garage it read 12V. So I tested another plug in the house: 0V. And another: 0V. I pulled the garage multiplug from the wall plug, tested this plug and it read 0V. I plugged the multiplug back in, tested and got 0V. I don’t know where the 12V went but can’t get it back. (I tested live and neutral to make sure my multimeter is working and it read 230V so that works.)
So this weekend I’ll be replacing the cable on the garage multiplug for peace of mind, but I’m not really closer to finding the culprit. (Wink wink, expert with expert tools needed.)
As soon as the fault gets wind that you’re onto it then it goes underground until next time… 
Which RCD did you install?
We’ve used the “higher immunity” Si (“Super Immunised” is the marketing jingo) model from Schneider which I’m certain is the A9R61623.
Other installers swear by the ABB F202 A-63/0.03 AP-R.
Also worth adding that the above was not sufficient and on some sites, we have had to split the essentials into multiple sections - with one three phase site I think have 9 earth leakages on the essentials.