I would express caution about cross-over because for both SARS-CoV-2 and the human genome you can obtain a good assembly from a reference genome. The SARS-CoV-2 genome is so small and read depth high enough this is not needed however. SARS-CoV-2 there are critical issues in assemblies for certain regions which have not been addressed, but for CDS there's no real difficulties here.
A reference genome assembly is not good practice for what are often environmental bacteria (bacteria per se) and de novo assembly is strongly recommended, even if it is to complement a reference genome assembly. Bacterial evolution is very different from either human or virus evolution, it's full of indels, missing genes, genes 'flying' around on plasmids and potential non-homologous recombination, risk of rearrangements etc ...
The challenge you face is automating the assemblies and then polishing those assemblies and there has been enormous progress recently, "closing the gaps" manually is a thing of the past - ancient history. There should be no under-estimation as the volume of recent work that has been dedicated particularly towards long read data. Even pre-assembly, pre-polishing the progress shouldn't be underestimated, albeit I suspect you're familiar with that. The old ideas of QUAST are now historic. The raised ante is noticeable.
Back to the original question consistently using a reference genome risks something going wrong that could cause an outlier artefact in the VCF and in my opinion its no longer cool for bacteria.
The rationale is clear, mis-assembly would cause an artefact in the tree and therefore in the epidemiological inference of the outbreak.
It's worth doing specific background reading because albeit I've deliberately not mentioned a given approach except for QUAST, the rate of progress has been significant. The world has changed.