I'm not committing to anything just yet, I'm just talking about my ideas. For example, if HTML5 performance gets good enough that an EXE exporter can only get a 10% performance edge, then an EXE wrapper around the HTML5 runtime should be perfectly adequate (possibly with a plugin to access Steam API features). Writing a new exporter is a massive amount of work which will tie us up for months and require a lot of continuous maintenance work, which is a difficult decision when you're a small team with limited resources. So the EXE exporter would have to be significantly better to justify the effort, which is why if it's at all possible I'd prefer to recycle the HTML5 engine.
Still, I think it's unlikely HTML5 will get that fast - I do think an EXE exporter for super-hi-performance games would still be a good feature.