Question: optimization: disappointments again

As usual happens in symbolic calculations - small change of input parameters leads to drammatical change in consuptions. I had talk about many other partial issues in  this forum. Many other issues i just tried myself without any success (for example try convert(exp(x)+x, FormalPowerSeries)) and now i must take decision: wheather or not to continue work with maple at all. It's totally clear for me that others (like Wolfram mathematica) not much better. Only one other case to solve my problem - implement series coefficients calculation by myself too.And then Jean-Baptiste Emanuel Zorg will 1001th time right again! "Want smth done? Do it yourself!"

For this moment i have code and all i want that functions will calculate fast enought irrespectivetly of input values (that consist over wide set of distributions and numerical values for them). Was partially succeded with k=1 before and now totally fail with k=3/2. Goddem what about other changes? Only series in general case should be solution. But how?

All i need for now - is implementation of 4 procedures:

IntermediateCalc

GetLimitsArray

SmallFuncSimpl

BigFuncSimpl

to get both reasonable result under CodeTools:-Usage namely for memory <200kb and for time less than 50ms. I'm pretty sure that this task completly solvable manually in all examples i should consider, but i don't know about universality.

Attach: main.mw.

All i want to know today: wheather it's possible to calculate so? YES or NO?