KD-Tree traversal (raytracing) - am I missing a case?

Just in case someone's interested - the mistake i did was to not consider a special case described in this paper

http://www.cs.utexas.edu/ftp/pub/techreports/tr88-07.pdf page 12

It happens if one polygon lies on the splitting plane, such that it's part of both cells and the ray goes through both cells. if the nearcell is tested, but the actual intersection happens in the space of the farcell (this is possible because the polygon that intersects is part of both cells) then there is still the possibility, that in the far cell a intersection could be found which is actually closer than the already found one. Therefore - if the found t for the intersection is larger than tSplit, then already the farCell has to be tested