It's recommended to do RUV correction before DESeq2. However, the PCA doesn't show any clear separation between the groups. How can I know if my data needs further pre-processing before DESeq2? Is there any metrics or plots that can guide me?

Check the RUVSeq package, they do have quite a lot of documentation about selection of number of factors.

For our work, I tend to follow their guidelines and also I look for association of estimated factors with things like processing date or known batch.