Known bugs and problems:

versions up to 2.52 beta

• potential array layout confusion - S2HAT routines permit two different layouts for the arrays storing alm coefficients. (See here for details.) The layout choice is decided by the input parameters called lda. Formally whenever it is equal to a number of components (e.g., Stokes parameters) considered the adapoted layout conforms with the HEALPix one, if lda is equal to lmax the layout is the native S2HAT one. In an unlikely even when the number of components is equal to lmax this convention may lead to a confusion. Note that routines do not check if the confusion is indeed possible but usually choose in such a case Healpix as a default.
• SOLVED in ver.2.54 !!! problem with gfortran compilers - some compilers seem to handle the F90 pointers differently than others. As a result though no compilation error is reported, the code usuing the library crushes. This seems to be related to the fact that some compilers for the indexing of F90 pointers to conform with standard Fortran convention, i.e., they have to run from 1 to vector size, independently of what are the index ranges of the associated array. This error has been found in a few gfortran distribution and is absent in others. It does seem that such behvior does not conform with the F90 lanaguage standard. Therefore, at least for time being, no attempt to correct for that has been undertaken. If the problem persists though, or I get convinced that this not a bug on gfortran side, that may get finally fixed. If your gfortran fails for the reason as described above, try to use some other compiler. At least for time being. Any suggestions and opinions on that are appreciated.