To understand how WGCNA treat correlations, have a look at this blog post from Peter Langfelder (WGCNA dev): https://peterlangfelder.com/2018/11/25/signed-or-unsigned-which-network-type-is-preferable/

Does that mean if in certain individuals (eg a disease state), Gene A goes up while gene B goes down - this counts as an expression pattern and you might find them in a given module?

If you are building an unsigned network, GeneA and GeneB will be included in the same module.

if I do WGCNA on say a group of patients with a disease eg rheumatoid arthritis and one module correlates with gait speed - does this mean that biological process in that module (eg if did overrepresentation analysis whihc showed eg erythrocyte development) is linked to gait speed in this population of rheumatoid arthritis patients?

Not really. Keep always in mind that WGCNA modules are “clusters of co-expressed genes” not clean biological pathways. Genes that overrepresent specific functional categories within a single module may have low module membership and thus distort the biological significance of that module. If you are primarily interested in gait speed, you should focus on module genes that significantly correlate with gait speed (high gene significance) and are also strongly correlated to the module eigengene (high module membership/kME).