Regression with SECTIONs

## Description

I have a problem with some cases where SECTIONs are used. It looks like in addition to running each of the sections there is one run that goes through neither of the sections. This wasn't the case in Catch 1 and it also works on most systems on Catch 2, but on one system I see test failures because of that. The only difference I can see so far is that failing system is a 32bit Linux, everything else is 64 bit. Maybe I can dig some more, but wanted to raise this here in case somebody has an idea what it could be.