I think simple tools that allow one to perform very basic bioinformatics type tasks may be those least appreciated from a sustainability point of view. Say bioawk or seqtk very useful but not clear how one could support them or publish them.
yet we need simple tasks all the time , how do I easily reverse complement a sequence, or translate it from the second reading frame, or find out how long a fasta file is without having to write a program in another programming language and then having to worry about how to install it which platform it will run on etc.
I've been using python statsmodels extensively lately. It has a very nice interface using R-like formulas via patsy. It has OLS, GLMs, mixed-effect models, GEEs, Robust linear regression, clustered errors, methods for calculating power/sample-size and much more.
It has good documentation and a very helpful mailing list.
Vmatch is definitely underrated. One problem is the manual doesn't really describe the Vmatch tools as they apply to common use cases. You'd really have to use a tool for one task before you would even consider using the rest of the toolset.
Don't get me started on overrated tools.
Lately I've been playing with SymPy, a python library for symbolic mathematics. It's still in development but it looks very nicely done and the mailing list is very responsive. I think it would be great if the bioinformatics community took it up.