The history of good models for spectra might be an example of missed discoveries. To briefly sketch this history: Spectra (in the sense of topology) were introced by Lima in a dissertation under the direction of George Whitehead in 1958; in the year 1964, Boardman gave a definition of the (homotopy) category of spectra and he also defined the smash product of spectra. But in his language, one could only formalize ring spectra up to homotopy. 
Peter May gave in his 1977 book *$E_\infty$-ring spaces and $E_\infty$-ring spectra* the first definition of an $E_\infty$-ring spectrum, i.e. a ring spectrum whose multiplication is associate and commutative in a homotopy coherent way. It was only in the 90s that people found models of spectra, where one can define $E_\infty$-ring spectra just as commutative monoids in a suitable category of spectra: Elmendorf, Kriz, Mandell and May came first with their $S$-modules and shortly after Jeff Smith defined symmetric spectra; shortly after that, Mandell, May, Schwede and Shipley defined the closely related model of orthogonal spectra. Compared to $S$-modules and the earlier formalizations of $E_\infty$-ring spectra, symmetric and orthogonal spectra are rather easy to define and produce a theory that is much easier to digest than the older ones. 

The interesting thing is: Peter May already defined commutative orthogonal spectra in his 1977 book under the name $\mathcal{J}_*$-prefunctors. He did not realize at this point though that they are the commutative monoids in a symmetric monoidal category of orthogonal spectra (and that this category has a homotopy theory that is equivalent to that of spectra) - reasons might be that the Day convolution product wasn't really known then to topologists and also that the language of model categories wasn't widespread. It seems that May viewed these $\mathcal{J}_*$-prefunctors only as a convenient technical input to construct examples of spectra in his language. 

In the 80s, Gunnarson (1982) and Bökstedt (1985) considered symmetric ring spectra without calling them this way or again without realizing fully their significance. 

Schwede's [unfinished book project][1] has history sections, which give more information about this. It is certainly fair to say that the history of stable homotopy theory and algebraic K-theory would have been less convoluted if people had found symmetric or orthogonal spectra earlier -- but this history should also highlight that realizing that these definitions were so significant, was a highly non-trivial insight by Jeff Smith. 


  [1]: http://www.math.uni-bonn.de/people/schwede/SymSpec.pdf