> What happens to the indices for the former Japanese sentence?

It gets left behind. I manually copied it to the new sentence in this case, but Trang will need to clear up things. I'm not sure if I can delete index entries.

> Should user be able to do this?

It's probably a bad idea.

Suppose you have

A(English) translates to B(Japanese) translates to C(German)

If you change the Japanese to a new, Dutch, sentence you get

A(English) translates to B(Dutch) (doesn't really) translates to C(German)

Because C is really the translation of the (vanished) Japanese sentence not the (new) Dutch sentence.

I think it would have been best to have left the Japanese duplicate and added a "Please delete me" comment.

> What happens to the indices for the former Japanese sentence?

As Paul said, it gets left behind. We don't have (yet) strong mechanisms that would help keep the database consistent.

> Should user be able to do this?

No, they shouldn't. Ideally, there should be guidelines (which I'm hoping to be able to write by the end of the month) to help users understand better how things work and how they can contribute in a way that doesn't give us (developers) more work than we already have ^^'
I wrote down some of the ideas in my comment here : http://tatoeba.org/eng/sentences/show/126

> Perhaps "nominate for deletion" could be added as explicit functionality?

Yes, in general, we could have various status for a sentence. Actually in the previous version of Tatoeba we used to have that, but I haven't re-implemented it. A sentence could be marked as "to delete", "checked" or "locked" (and perhaps other things, I don't remember). When a sentence was checked, it meant you could rely on it for not having mistakes. When it was locked, no one could edit it anymore.

But this is not urgent compared to other things we have to do. My priority at the moment is to make sure that people understand clearly that when they translate, they have to translate from the sentence written in big letters. I'm pretty sure that very often, people are adding translations to a Japanese sentence when they were actually translating from the English sentence.
We also have to enable people to link and unlink sentences. There are many sentences that are linked to each other without being translations of each other, and there are many sentences that could be translations of each other but are not linked to each other.

Once all of that is settled, and people understand that they have to view the corpus as a GRAPH and not a table, it will be less likely that they behave in a way that we don't want them to behave, like what you did. And perhaps "nominate for deletion" will not be *that* useful because instead of deleting, you could just edit your sentence into whatever you want and unlink it from any sentence it was linked to.