Depends on what you scaling. Actually there is no borders in such approach. If you app overloaded with template parsing you could move it to another machine, if problem in "database" thing same solutions came as for NoSQL. There is no limit what to scale and how. You could have everything in one process exept something that need scaling and this would be more efficient than dividing things on databases, cgi and balancers.
In case of using it as library - if you write some application you use different api to make things work, for example call api to write info into file. But, if you face with scaling or other problems in most cases you need to rewrite code of your application, for example set up buffer before writing to file. If you use frozen you can define format in which you dump info to it and stop touching code and binaries of your application at all. Quick changes in configuration file to add buffer, or write into memory, or flush info by network and everything is working.
Feel free to fork it on github and make contributions. To avoid double work, check out TODO file - top items have first priority and would be done very soon.
There is one nice example for now - split file on many using set of regular expressions. But i get it, will write more examples including web application, caching and so on.