I encountered a weird relax-optimization output in a structure containing sugars. The output structure has hydrogens in close contact (~1.2 Å) from a sugar molecule and a serine residue. In this reported case, I use the -include_sugars flag, but I've observed the same behaviour with custom params generated for glucose in another system, in which I found hydrogens in close contact with a tyrosine. It seems to be an issue of the Hpol atom type, which allows this to happen for a hitherto unidentified reason. The attached relaxed PDB (truncated at the C-terminus to fit the 512kb limit for uploading, which is a bit low for PDB if you ask me) shows the interaction between hydrogens from residue GLC5 (chain B) and SER105 (chain A).
Any clue on this behaviour will be appreciated.