It's not a huge problem, as long as Spectre does eventually find a solution ... it might be taking longer than it should to find that solution.

I think Spectre tries to make a "good" initial guess for DC Newton -- a "dumb" initial guess would be: all nodes are zero, except those that have voltage sources on them.  One could attempt to improve upon this by using some heuristics to guess some other values: if you know the voltage on one side of a diode, then the voltage on the other side might be 0.6V higher.  I think the message is saying that Spectre's heuristics are having trouble with your circuit topology or some of the device models.