On LU-3245 Andreas said "Moving forward beyond the 6209 patch landing, it makes sense to explicitly block rename for the FID_SEQ_DOT_LUSTRE objects, since I don't think it ever makes sense to allow that."

We should also review the other reint handlers checking for consistency with respect to the weird FIDs and, I hope, factor the cases into one or more "client may access this FID for R/W/X/..."