All right, I'll bite: see Retrospective laws breach separation of powers.
A bill of attainder is an obvious breach of the separation of powers. A retrospective law creating criminal liability is tantamount to a bill of attainder.
Furthermore, because a judicial finding of law is retrospective to the last relevant legislative change, a legislative change cannot be retrospective without usurping a judicial function.
Even without an explicit separation of powers, the mere existence of a constitution, or of a court, recognizes the rule of law, which in turn gives an implicit separation of powers.
The rule of law has three further implications which exclude retrospective laws creating criminal liability. First, the law must be knowable. If a retrospective law is to satisfy this requirement, it must be so obvious that it is knowable before it is enacted. Second, the law cannot demand the impossible. Hence it cannot require us to know a law that has not yet been made and is not knowable until it is made. Third, we are under a "government of laws", but a general power to make retrospective laws would place us under a "government of men".
It is arguable that a retrospective law is permitted if it only produces an otherwise permitted effect. At a stretch, that argument would permit a retrospective law criminalizing activities that were already crimes under non-retrospective laws (though one might ask what is the point). But it does not support retrospective legislation criminalizing activities that would not otherwise have been crimes at all.
(e.g. IANAL; TINLA, etc.)
P.S.: Response to Samuel Russell's comment
Samuel Russell writes: "Putland's analysis in that blog post relies on the US constitution to make an argument in relation to Australia..."
Not quite. My discussion of the U.S. Constitution in the blog post concludes that "bills of attainder and ex post facto laws..., if they were not explicitly forbidden by particular clauses of the U.S. Constitution, would be implicitly forbidden by the separation of powers." And of course Australia has the separation of powers. Hence I didn't find it necessary to mention the U.S. constitution in the above summary.
While opinion varies on how much is implied by the rule of law, the minimalist opinion is that the law must be knowable and therefore prospective. If that's all it takes to be "tendential", so be it. My contention that the rule of law and its implications are already binding, and not merely aspirational, is more original; but Russell doesn't argue with that.
He does argue against my claim that the rule of law implies the separation of powers. But the rejection of ex post facto laws survives without that claim; it follows from the rule of law, and it follows from the separation of powers, regardless of any connection between the two.
Concerning the UK, I admit that the separation of the Supreme Court from the House of Lords was centuries overdue. The former arrangement did not give a structural separation of powers. Whether it gave enough functional separation to satisfy the rule of law depends on the effectiveness of mechanisms for avoiding conflicts of interest (on which I'm not in a position to comment further). Note that my derivation of the separation of powers from the rule of law does not specify the extent to which the executive branch is answerable to, or drawn from, the legislative branch; thus it neither endorses nor rejects the Westminster system per se, although it does warn against tight party discipline that results in a domination of Parliament by the Executive.
Concerning the history of the separation of powers in Australia, I am aware that the High Court has taken a narrow view of what tribunals can exercise judicial power. But I would have thought that strengthens the case against usurpation of judicial power by the legislature.