Hi @Siyu_Song ,
This is a really great question. The short answer is no, there should not be an issue, but as always it depends.
It depends on whether you will need to directly compare results generated by these two resources. In a standard 16S workflow the phylogeny is usually only used for diversity analyses, and the ASVs (not the taxonomy) are used as tip labels to relate this to your feature table. So the different database versions should not cause an issue.
This should not cause any incompatibility issues, but indeed the SILVA 128 release is quite a few years old and so might not be as complete as the latest version, so there is a small chance that you might fail to insert (and hence filter out) some ASVs into the tree. But this seems unlikely, as the point of using SEPP (vs. a tree mapping approach) is to insert new sequences into a tree.
No I think not, because you are just using the SILVA 128 tree as a backbone to splice in your own ASVs, and the tips are then labeled by those ASV IDs. I suppose if you want to annotate/color the tree by taxonomy this is where inconsistencies could cause issues (as the output tree would contain both your ASVs and the reference sequences), but I am not sure.
I hope that helps!