Separating abrasion and corrosion is moot, in my opinion, so when we say "corroded," I would take it to mean "abraded/abrased" as applicable.

If you want initial stresses, set the corrosion allowance to 0 for the model.

For end of life, use initial thickness with corrosion allowance.

Yes, it uses the strength of the thinner material and the weight of the thicker material. We're not doing this because it's "unfavorable" and because it's "conservative."

We're doing this because we cannot predict with certainty that any given point in the system will be the first to corrode and everywhere else is as thick as it was installed.

If you're measuring abrasion through the life of the pipe, there's nothing stopping you from creating a "current day stresses" model with present day thicknesses and 0 corrosion allowance.

There's also nothing stopping you from creating "next year's projected piping at present rates."

And the year after.

And the year after.

It'll be up to regular inspection to verify your predicted rates.

It'll be up to you... or your successor... to adjust the analysis accordingly as new information makes itself known.

With respect to SIFs on fittings, if you want to get the most accurate handle on stresses, put CNODED anchors on the fitting welds, and model it in FEA, again, attempting to predict the future based on the info you have now.