It seems that when proofs are replayed, formulas built with different formula factories are getting mixed up. This caused some replay bugs in the Theory plug-in. They were fixed by translating the formulas in each reasoner of the plug-in, but it would be better to handle this properly in Rodin core by translating formulas to the right factory before providing them to the reasoners.