I’ve built this filter, in the Kosmo version by @sebastian, and can verify it self oscillates down to about 135 Hz. Not very large amplitude at that frequency, but it’s certainly there.
The Kassutronics instructions point to these directions for doing the transistor matching. I wrote up the process and a stripboard layout for the matching circuit here. But it seems to me unlikely transistor matching is the problem here. More likely a component or connection problem in the feedback section, I’d guess. And it’s conceivable the breadboard connections might be somehow limiting the gain at low frequency.