Cell type sub-clustering across multiple samples in scRNA-seq

Question

[narges.bioinf]

Hi,

This still might be controversial and also project-specific question, but under what circumstances one should or should not apply integration/batch effect correction when focusing on one specific cell type from different samples? Note that the integration and clustering for all cells has been done before and now we are dealing with sub-clustering of a specific cell type from the scratch.

Thanks!