I don't see why an infinite regress is intrinsically impossible, though II don't think real infinities exist, but as an alternative to it and to a first cause, how about a loop, eg a causes b, which causes c, which causes d ... which case y, which causes z, which cases a, which causes b, etc ?