Let's start with your definition.

"Selection for traits that would be beneficial to a population of units at the expense of an individual unit possessing the trait"

This is not a good definition of group selection. In reality, selection can act on groups regardless of the direction of selection at individual level. This definition sounds to me (your link is directing me to somewhere unpleasant) like one that has been set up in order to study whether group selection can act as an explanatory principle for the evolution of individual-level altruism, which is an altogether different question than the title of your post (the two questions are often conflated for historical reasons).

To expand the debate, I will introduce two formal definitions of group selection both of which are defined by alternative forms of Price's equation (See Okasha's book Evolution and the levels of selection for a thorough analysis).

Interdemic group selection

Interdemic selection occurs when selection acts on both groups and individuals within groups separately. That is, both groups and individuals within those groups must be units of selection in their own right. We must be able to assign fitness to both levels of selection some way, and track the offspring produced at each level. In this form of group selection, group fitness corresponds to the ability of a group to survive and produce more groups of a similar kind, and individual fitness corresponds to the ability of an individual within a group to survive and produce more individuals of a similar kind within that group. A good example is selection on multicelullar organisms and selection on cells within those organisms. If there are greater than two levels of selection, interdemic selection is commonly referred to as multi-level selection 2 (MLS2) (see Okasha).

'The extinction of some groups and the ``reproduction'' of others are essential features of evolution by group selection. If groups are the units of selection, then they must have the properties of variation, multiplication, and heredity required if natural selection is to operate on them.' Maynard Smith

Intrademic group selection

Intrademic selection occurs strictly among individuals within a meta-population. That is, individuals are the only units of selection we study. The groups in which individuals reside are seen as part of the structure of the environment in which individuals live. In this form of group selection, individual fitness corresponds to the ability of an individual to survive and produce more individuals of a similar kind, given that they reside in a particular group. A good example of this perspective on group selection is selection cells of the quasi-multicellular organism Dictyostelium. Cells of this species are units of selection in their own right that occasionally group together into bodies that may also act as units of selection. If there are greater than two levels of selection, interdemic selection is commonly referred to as multi-level selection 1 (MLS1) (see Okasha).

'A way of looking at this is that the [intrademic] group selection approach looks at the evolution of individual characters in a group structured population, whereas the [interademic] group selection approach looks at the evolution of group characters.' West et al

Equivalence

It is worth pointing out that there is currently equivocation in the field about whether these two selective processes are different perspectives on the same underlying biology or distinct causal processes.

Another important point to note is that it is the intrademic form of group selection that leads to the conclusion that kin selection and group selection are really just different perspectives on the same selective process. This has been shown mathematically by Queller.

There is not yet a conclusive proof either way showing that interdemic group selection and kin selection are equivalent or not. Some evolutionary theorists feel that they are, while many feel that they are not.

Group selection versus group adaptation

It is also important to point out the distinction between the process of selection acting on groups (which most evolutionary theorists accept - I mean, it definitely does happen!), and whether or not this selection gives rise to group adaptations (which evolutionary theorists are equivocal about). Some argue that all adaptation can be explained from selection acting on lower-level units (right down to the gene, remember Dawkins?), whereas others think there may be nested levels of causality that cannot be reduced in this way.

Empirical research

There are many examples of laboratory and field based studies of group selection. Too many to list here. In a response to a theoretical paper by Gardner, a history of empirical research on group selection was given by Goodnight. The original paper, its response, and its counter response are worth a read to understand some of the different perspectives on the debates. In particular, two different perspectives on the evolutionary dynamics of cancer are presented that nicely illustrate the differences between MLS1 and MLS2 thinking.