I'm not sure about validity but it will surely make assembly a lot more difficult (MAGs are a great thing but really knowing this is from one bacteria is better). You will basically be creating a bacterial pool. The reads processing would be similar to metagenomics sequencing? Which is more difficult than assembly from WGS because it's a mix of genomes. Do you plan to use Illumina? Or long reads? And if you pool more bacteria together you will need a greater depth of sequencing (and there is the possibility that some will be less represented and the assembly will be poor for some bacteria). There are a lot of pitfalls but I agree that would probably be cheaper. But this is just a few thoughts you should consider. I'm interested in the answer myself.

In addition to what Dominika Kadlečková said, I would add that it depends somewhat on how similar the bacteria are. If they are in the same family and can exchange any plasmids or phages, then the assembly may become uncertain due to identical regions in two or more of the chromosomes or plasmids present in the isolates.