If the password complexity is small, than sure, you can use rainbow tables. But if the password get longer and longer, than have do you want to store the rainbow tables? Also if you have salted passwords, how are the rainbow tables going to help?

Not really. Rainbow tables are inflexible TMTOs that are slow to construct and have been made largely obsolete by advances in GPU cracking and more widespread use of good salts. Your WPA2 example doesn't hold up in the real world, where even default manufacturer SSIDs typically append a random number to avoid collisions with neighbors. One common pattern is manufacturer_####, that would be 10k rainbow tables per manufacturer and would only cover the defaults!

Honestly if I hear someone in infosec start talking about rainbow tables in 2022, that's a good indicator that they have not cracked passwords in a long time, or they are just regurgitating what they learned to pass some basic security cert. Hashcat on a modern GPU is blazingly fast, and barring some really niche edge case, your best bet is just going to involve throwing more GPUs at the problem.