Well, not really... but it should be avoided for one obvious reason: Delays in a potential switch will only make the porting take longer.
I can only speak from experience in regard to this android project of mine - I first started out in AndEngine due to the fact that simple things were simple, with ready-made libraries for most of what i wanted to do. However, with time, the effective speed of the project was slower and slower due to AndEngine's complete lack of documentation, coupled with some performance issues. In the beginning I was doubting whether i had chosen the right engine, and I came across libgdx as a potential replacement. However, silly me concluded "I'll get this and that feature done and get the next alpha version up and running before I'll test with a different engine".
My conclusion: If in doubt, try swapping the engine as fast as possible. It also enables you to rewrite parts you perhaps approached the wong way to begin with, from a design-perspective