Glad to hear of your interest Thom. I've renamed this thread to align more closely with your question. I'll expand my response to address some related issues.
Time-varying growth parameters, including cohort specific growth have definitely been used. We have no way of recording how SS3 is used in practice, but I certainly am aware of such uses.
Multiple growth patterns have also been used; mostly in context of a multi-area model that needed different growth pattern for a northern vs southern morph.
I am not aware of anyone using the platoon feature in an assessment. Platoons subdivide each sex x morph into multiple platoons, not with different growth parameters, rather by simply breaking the overall length-at-age distribution at birth into a range of growth trajectories such that the larger platoons experience more cumulative effect of length-dependent fishing mortality and causing the average length-at-age of older fish to decline because of this differential mortality.
Another available feature is seasonal settlement. The total recruitment can be distributed into multiple settlement events, each with a unique date on which they enter the population and begin growing. Even though all of these settlement events may have the same growth parameters, early born fish have a head start so at a particular time of year, they are larger than late born siblings. So, this settlement phenomenon, which must happen in nature for any species with an extended spawning/recruitment season, can produce an effect just like that of platoons.
An interesting aspect of settlement spread is that the spread of length-at-age across the group will be greatest during the fast-growing young ages, then converge as they all grow towards the same Linfinity. I believe this phenomenon is also observed in actual data. Look for declining std.dev. of length-at-age as fish get older.
Finally, investigating phenomena like above takes detailed data. I do not think a good investigation of growth can be done with annual data. I highly recommend seasonal time steps and paying close attention to exactly when during the year samples are taken. SS3 needs to know exactly what time of year to calculate the age-length key for. This is why survey and composition observations are entered with a time stamp of decimal month, not season. SS3 then assigns each month entry to the relevant season and calculates elapsed time from start of that season to the time of the observation. There is one more step to refining the match; this is the subseas feature that you enter near top of the data file. Normally people use just 2 subseasons and SS3 calculates the ALK at beginning of each subseason, which is beginning and middle of the season and which may be too crude to get accurate alignment between the model and the length modes in the data. When you enter a larger value for subseasons, SS3 is prepared to calculate the ALK for each of those subseasons, but it only does so if there is an observation that is timed close to that subseason.
I suppose a final thought is that surveys, with a duration of 1-2 months, should provide more precise information on growth than fishery samples for which samples from a wide range of moths are accumulated into a simple "sample". SS3 does not yet have a correction for this, but I am thinking about it. Potentially you could use the "super-year" feature to cause SS3 to calculate the expected length composition for each month of a season and then combine those into an overall expected value to compare to the observation collected over the same range of months. That would be quite a detailed model, would be big and run slowly due to all the computations, and probably much more suitable for a research project regarding fish growth, not an operational assessment.
Rick