Almost all modern symmetric encryption algorithms consist of multiple "rounds" of a similar sub-algorithm. Sometimes they have special operations at the beginning and/or end of the process, but most of the work consists of repeated iteration of more-or-less the same simpler sub-algorithm. Each round performs a bit of encryption all by itself, but the bits typically become even more diffused with repeated application of the sub-algorithm. In some cases, rounds are slightly different from each other in the sense that they are indexed by different key-derived values or the like. But usually the gist of the sub-algorithm remains the same.

Often cryptanalysts begin attacks on an algorithm by attacking a "simplified" version of the algorithm that has fewer rounds. Well-tested algorithms have a very carefully chosen number of rounds. It is rare that adding more rounds will weaken plausibly strong algorithms. But one thing that adding extra rounds always does is add more computational burden to performing the encryption. In practical uses, you always want a faster algorithm rather than slower one, all other things being equal. So the goal in designing an algorithm is to have enough rounds to make it secure while having as few rounds as possible to keep it fast.

Of course, extra rounds added to a bad starting algorithm will have a limited effect. For example, the Gnosis cipher presented above has a rather undesirable property when it comes to rounds. Performing multiple rounds of the Gnosis cipher is equivalent to performing just one round using a different initial key. Adding rounds has no effect whatsoever on the strength. If this is not immediately obvious, it is worthwhile to page back and review the Gnosis cipher in order to understand why this happens. The effect is similar to, but simpler than, problems and limitations encountered by earnest attempts at creating encryption algorithms.