If the time points are different, you should not merge them. At different time points, many genes are expressed differently, hence you should not combine them.

This depends a bit on the type of analysis you're doing. These will likely differ from your samples due to being prepared elsewhere, so that could throw things off. It's likely that you'll want to do a PCA or run things through SVAseq or Combat.

My point is that given that both experiments have enough time points the most robust conclusion would be the one that was replicated separately in both time courses. Combining datasets could increase the statistical power of the analysis, yet the major drawback could be possible batch-effects if those experiments are not properly normalized.