Sometimes a solution is complex and when confronted by this, many people follow their innate desire to find a more simple and "elegant" solution. This of course is a noble endeavour. A simple and elegant solution is probably less risky to schedule, cost and resource. It is probably far more cost effective to operate and maintain. It is also probably more available, scalable, highly performing, secure and so forth.
The counter point to this is perhaps nicely summarised by Albert Einstein (thanks to colleague Sid Kargupta for pointing this out): "Everything should be made as simple as possible, but not simpler". To architects this reminds us that some solutions are inherently, fundamentally and irreconcilably complex either in part or whole. Attempting to make these parts any simpler will mean that the solution begins to lose something. It is better to recognise and respect the nature of the problem than to undermine it.
To quote someone quite different to Albert Einstein: "Shoot the hostage". If you remember Keanu Reeves in Speed you'll know what I am referring to. The character in the film essentially, albeit unusually, eliminates the hostage situation allowing him to pursue the criminal. In the architect's case, when presented with a solution that cannot be further reduced, it is not a bad strategy to go back to the client to re-examine the requirements that culminate in the complexity. If you can shoot the hostage it might be better all round. Does the customer realise the ramifications of their requirement? Was the requirement born of absolute necessity without compromise or was it a reasonable expectation given the understanding at the time?
A classic case of this is the "the solution must be five nines availability" demand. Some solutions will have a hard time solving this problem full stop. Add the cost of such a solution into the mix and often it is not commensurate with the risk and exposure the requirement is seeking to mitigate. The requirement is reconsidered, giving scope to accommodate different solutions.
So, no, complex architectures are not wrong as long as they are simple as they can be (no simpler) and that the requirements that culminate in the complexity have been thoroughly validated in the new light shed by the implications.