The problem is split into smaller tasks, and many computers work on those tasks at the same time. Their results are then combined to solve the larger problem.
Not necessarily. Computers need to communicate and coordinate with each other, which takes time and creates overhead. Also, some tasks cannot be easily divided between many computers.
Yes, if they use a system designed to handle failures, such as Byzantine Fault Tolerance (BFT). The number of computers that can fail or lie depends on the rules of the system.
Most of the heavy computation happens in remote data centers containing many powerful servers. Your phone or computer sends requests to these servers and receives the results.
They could handle extremely complex tasks, such as detailed simulations of the human body, climate, ecosystems, and entire virtual worlds. They could also help optimize global systems like energy grids and supply chains.
Supercomputer — Wikipedia
Distributed Systems — Encyclopedia
Agreement Protocol in Distributed Systems — GeeksforGeeks