When it comes to the discussion of Gabor systems the focus is often at the identification or verification of pairs (g,Lambda) which allow to build a (regular) Gabor frame from a Gabor atom g by moving it through the time-frequency plane along the lattice Lambda.

It is much less popular, although probably more relevant for applications, to discuss another set of questions, namely:

1) Under which conditions can one be sure that - despite moderate amount of redundancy - one has a good Gabor system
2) How can one find suitable lattices for a given window (e.g. a generalized Gaussian)?
Of course this is related to (sub)optimal sampling strategies for the STFT taken with this window.
3) Given the lattice of sampling, what may be a good window.
4) In applications it may be of interest to choose the Gabor system which optimally allows to
approximate a given slowly varying channel by a well-chosen Gabor multiplier;
5) When one talks about Gabor systems, one often wants to see (for good reasons) well localized dual Gabor atoms. This can be seen as yet another optimality criterion.

In this talk we will suggest a view results, partially in terms of experimental evidence gained during the UnlocX project, partially indicating which concepts should or can be used in related situations, where Bessel sequences in Hilbert spaces are used for signal processing applications.