I feel like it should be noted that while it's useful to understand this exact algorithm, processing that entire algorithm afresh is of course not how words are produced or understood at conversational speed in normal sentences. The way the rules are enacted at speed is (as with many things in language) through broad fast pattern recognition. The rules produce numerous patterns, each of which gradually become increasingly familiar and then obvious and then reflexive.
For instance one of the particular patterns implied by the full set of rules is that you can put together any two CCV rafsi to make a CCVCCV lujvo like {brivla}. Without understanding all of the implications and edge cases of the whole set of rules you can simply take this example and start producing that shape of lujvo with whatever CCV rafsi you know. How about CVV and CV'V rafsi then? Those work there too, CCVCV'V, CCVCVV, CVVCCV, CV'VCCV, all good, the only complication is with CVV + CVV you'll need an -r- hyphen (or very occasionally an -n-), which makes sense if you compare with how CVV CVV is two cmavo. Knowing a few shapes like that makes a zillion combinations you can make, you can start making lujvo all day long, and then when you learn which CCC are ok you can feel confident making CVCCCV or CVCyCCV shaped lujvo too, etc.
Similarly you don't need to know every zi'evla shape in the universe to start making zi'evla. Every word you know is an example of which forms are allowed. If you learn the word {sorpeka} (bus) then you've also learned that you can make any word of that shape, like IDK, {bamnosa} or {morpuki}. I don't need to compare {morpuki} to some complicated algorithm-- I can compare it to {sorpeka} and see that it's the same shape so it's allowed and it's in that category.