There've been quite a few replies explaining that this measurement isn't
valid, or at least is irrelevant to the problem, but I am note entirely
clear on why. My understanding of the test that you are performing is as
follows:

Appliance L connected to Supply L
Appliance N conencted to Supply N
Appliance E - not connected
Supply E - not connected

At last, someone who can see my point )

You then place the multimeter from the appliance earth to supply neutral,
and it read 2.9A. To me, this clearly doesn't sound healthy. In a
healthy appliance, the earth should isolated from both live and neutral
(with the exception of a high impedance link via the noise suppression
caps), and should hence not be able to supply any current to neutral.

....yeeessss!

The point that was being made is that if you place a multimeter between
the supply earth and neutral then significant current will flow. This is
clearly true, as N and E are held different potentials via a relatively
low impedance, depending on the earthing system. However, in Sparks'
case, he is not measuring current from supply earth to neutral, but from
the appliance earth to neutral, and the appliance earth should be isolated
from any potentials as he has disconnected the earth terminal. It is
possible that some form of supplementary bonding is in place, and
introducing a potential on the appliance earth, but this seems unlikely
because his RCD ceased to trip when the earth terminal was disonnected.
Also, the fact that the RCD doesn't trip, even when the multimeter is
showing 2.9A flowing to neutral, suggests that this current is being
sourced from the RCD protected supply.

If my understanding of the tests that have been performed is correct, then
this does sound clearly like a faulty device.

I eventually managed to get through to someone at Siemens who has a clue,
and it is going back!

Chris Key