The results are perfectly valid without the beta prior and this is how nearly all other RNA-seq software return fold changes (without a prior-based moderation).
If you read the DESeq2 paper, we describe how the prior on fold changes ("beta"), is useful for visualization and ranking, but it has little effect on inference (p-values, adjusted p-values).
For whatever reason, the prior could not be estimated for some of your coefficients, so anyway there is not the choice.
So I'm not sure the origin of the error, but you can eliminate it just by setting betaPrior=FALSE when running DESeq().
For some reason the prior is not estimable for one of the coefficients.
and will this affect my results?
The results are perfectly valid without the beta prior and this is how nearly all other RNA-seq software return fold changes (without a prior-based moderation).
If you read the DESeq2 paper, we describe how the prior on fold changes ("beta"), is useful for visualization and ranking, but it has little effect on inference (p-values, adjusted p-values).
For whatever reason, the prior could not be estimated for some of your coefficients, so anyway there is not the choice.