Does anyone have a sense of the prevalence of use of gVCF vs. VCF by clinical sequencing labs? (My understanding is that the use of gVCF is growing)

I think the bigger trend is the slow death of joint variant calling favor of single-sample variant calling. Modern callers have enough existing data to no longer rely on batches to detect artifacts.

Also, does anyone know if labs are producing gVCFs using block format vs base pair resolution format?

When I see bp_resolution it's usually because the developer is trying to use home-grown or non-standard tools and doesn't want to go down the rabbit hole of parsing a gVCF with reference blocks.