Possible analytical approaches:
Hermisson/Pennings with moving optimum (selection gradient increases linearly); use results from Uecker and Hermisson 2011?
Hermisson/Pennings for quantitative trait (functional relationship between selection coefficients before and after environmental change; ask which mutation from sgv is most likely to fix first)
use bi-allelic infinite sites model without linkage as a simple extension of one-locus case
Concentrate on HC-regime (relatively low mutation rate, distribution of new mutations independent of parent allele; predominant wild-type allele)
Potentially useful papers: Keightly and Hill 1988, Foley 1982)
Simulations: