This all seems a bit vague to me tbh. Let's not add APIs because we could need them at some point but better when we actually know what we need and that we need it. I don't like bloating contracts with stuff which only is needed for some specific implementations.

The ES backend is the only one which might currently be able to implement this as a no-op.

Ok, but then let's add it now to DirectoryBasedIndexManager, we still can promote it to IndexManager down the road if actually needed. Also it seems it's mixing two concerns: "closing the index writer" for sure sound different from what's described in HSEARCH-2138.

Several other backends would benefit from being able to handle such a signal.

Yes, but they all extend DirectoryBasedIndexManager, no?