Now, taking from my "relativistic electromagnetic metric in medium" worksheet...
from the section "special relativity electromagnetic stress-energy in medium":

T(uv)=↓u(i)[12(1ϵD2+1μB2)121c(n2+1)Sj121c(n2+1)Si12ηij(1ϵD2+1μB2)1ϵDiDj1μBiBj]v(j)
Tuv has units kgms2

For electromagnetism, T00 and T0i are given by:
T00=12((ϵ)1D2+(μ)1B2)
T0i=c2(vpc)2ϵjklDkBl
Ti0=c2ϵiklDkBl
What do we do about the asymmetry? That brings on a whole other level of complexity.
In fact, if we are using the de Donder gauge then we are stuck with
12h¯ab=Gab
aka
12(hTR)ab=(RTR)ab
With 'TR' meaning 'trace-reversal'.
This means that any asymmetry in the Ricci curvature will also have to show up in the metric (right? or will the get rid of it? I can't see how...).
It does make me think I do, in fact, need to use a symmetrizing mechanism. Options?
There's the simple Tab:=T(ab).
There's the Belinfante and Rosenfeld symmetrization: Tab=Tab+12(sbac+sbca+sacb),c
But this means also introducing our torsion term sabc which should be coupled to our particle spins... 2T[ab]=sabc,c

if we simply symmetrize we get:
T0i=121c(n2+1)Si
For, D×B=ϵE×μH=1(vp)2S=1c2n2S, for S the Poynting vector.