The quality of large and complex Systems-of-Systems (SoS) that have emerged in critical application domains depends on the quality of their architectures, which are inherently dynamic in terms of reorganization at runtime to comply with domain needs. However, the way to model and evaluate the quality of these architectures is not clear. This article presents the state of the art regarding how SoS architectures have been evaluated. For this, we systematically examined the literature and, as a result, we discovered and summarized relevant architectural evaluation methods and associated modeling techniques and quality attributes, the maturity of these methods, as well as the benefits and costs of adopting them. We also address open issues and research opportunities and recommend that the mindset for SoS architecture evaluation must be changed to ensure the quality of SoS in critical domains.