This could very well be an implementation issue. sweights used to support only unbinned fits, as for a long time unbinned fits where the only possibility in zfit. It could very well be possible that, as binned fits became available, simply nobody thought of this option.

I would therefore suggest to head over to hepstats and report it there or better, suggest a fix. To try out what works, simply change it locally (it's probably an easy fix?) and suggest it in an issue, most likely a PR will gladly be accepted.