Hi Axist,
 Is this possible to combine both a dynamic and static set of tickers?
Yes, it is possible.Refer to the following thread. As we already have a dynamic universe adding securities manually may lead to survivorship-bias in the algorithm
If not, how would I go about using only the Coarse selector? I suspect using OnSecuritiesChanges and then from there?
We can store the list of selected securities with the help of the OnSecuritiesChanged event handler. Refer to the following code snippet.
 

 def OnSecuritiesChanged(self,changes):
     for x in changes.AddedSecurities:
       if x.Symbol not in self.symbols:
         self.symbols[x.Symbol] = x.Symbol
     for x in changes.RemovedSecurities:
       if x.Symbol in self.symbols:
         self.symbols.pop(x.Symbol)

Refer to the following thread for more information about using the optimizer.
 

What is the best way to test optimizers such as MaximumSharpeRatioPortfolioOptimizer or another?

 The MaximumSharpeRatioPortfolioOptimizer is designed to optimize portfolio weights using historical returns of the securities. To create a SciPy optimizer that optimizes the first two arguments passed to the RandomForestRegressor, a new optimizer would need to be created. Please  refer to this SciPy tutorial for more information.
We have updated the above algorithm where we use manual Universe selection using only the coarse function and keep track of the added securities in the OnSecuritiesChanged event handler. Note that we can increase the speed of backtest consistently by removing the History call again and again. We can store the historical data in a rolling window and keep it updating with the help of a consolidator. Refer to the attached backtest.
Best,
Varad Kabade