Build DAG is not generated properly

[c-mertes]

When building t-mae https://github.com/bioconda/bioconda-recipes/pull/31069 I had an issue with bioconductor-rhdf5. Fixing the recipe in the same PR (3 recipes got changed) still ended up with a failed build since t-mae was built first as you can see here.

I guess the problem is that bioconductor-rhdf5 is an indirect dependency of t-mae and adding bioconductor-hdf5array as direct dependency created the correct building DAG as you can see here.

Not sure where to fix this, but in my point of view, the dependency within the building DAG is not complete.