Many node-organizing blockchain models for IoT lack the scalability needed for high transaction volumes and mechanisms to manage decentralization and node overhead. This study proposes a multi-tiered Self-Scalable Dynamic architecture based on a forest of trees to improve scalability and efficiency by clustering all nodes to limit those involved in consensus. In addition to the architecture, our contributions include: i) A protocol to speed up consensus. ii) A mechanism for efficient node-lookup and message-routing. iii) A dynamic algorithm to balance overhead and decentralization. iv) A colocality mechanism to enhance consensus time. This architecture reduces consensus time with an upper bound of O(1) plus a propagation delay factor. As more IoT devices join, resulting in a deeper forest, the model becomes more scalable with shorter consensus delays. The proposed mechanisms may serve as general-purpose solutions for scalability, load balancing, node-lookup and message-routing, and are application-, clustering-, and consensus-algorithm-agnostic.