I am working on an analysis started back in 2018. We used the v6.12 of Root and we developed everything using this version. As a test I ran our fitting code in the newer 6.24 version, and while the fit converges normally, and the central signal yields are very close, the uncertainties are different. the v6.24 has a factor of 2 larger error. I tried to play with the RooFit options (like adding Minos etc) but do far I could not pin down the reason. I can try to create a toy code to reproduce it and send it here, but I thought good to start by just asking if anyone has an idea. Thanks.