How did you prioritize transcripts in the vcf? VEP provides options to order/pick transcripts (see the link below) by their consequences or relevance. It seems that no pick order was specified and transcripts are randomly written in the into gnomAD v4 vcf files. Since I am unable to find methodology for data processing, I wanted to asked here.
https://www.ensembl.org/info/docs/tools/vep/script/vep_options.html#pick