I'm playing with a simple example to learn how to use the relatively new ARProcess (and related functions) on the path to working with state-space models and filtering. Here is a simple scalar example that works as expected:

In[23]:= CovarianceFunction[ARProcess[{a},\[Sigma]^2],s,t]
Out[23]= (a^Abs[s-t] \[Sigma]^2)/(1-a^2)

I have a toy example for a multivariate state equation with Gaussian noise (in the finance world, it's an AR(1) model with Gaussian innovations):

In[21]:= MyProc = ARProcess[A, \[Sigma]^2 * IdentityMatrix[2]]
Out[21]= ARProcess[{{0,\[Lambda]},{\[Lambda],0}},{{\[Sigma]^2,0},{0,\[Sigma]^2}}]
In[22]:= CovarianceFunction[MyProc, s,t]
Out[22]= CovarianceFunction[ARProcess[{{0,\[Lambda]},{\[Lambda],0}},{{\[Sigma]^2,0},{0,\[Sigma]^2}}],s,t]

As you can see, the expressions pass the parser without error, but Mathematica doesn't produce anything useful even for this trivial 2x2 example. Any pointers to my errors? I used Mathematica 30+ years ago, and am just now returning to it.