As others have said, your thinking here is very much "apply the contract as if it were code, and then think of exploits/bugs". I don't believe this maps to the real world legal system very well.
Only if those two parties actually colluded with that intent, and you can prove it.
Personally, I don't want to be on the hook for proving this kind of nonsense in order to enforce my license. If I cared about SaaSification, I'd much rather use an outright EULA that forbids it and cut out all this nonsense of pretending you can solve the issue with a free software pure copyright license.
Let's think about it though -
Party 1 hosts the software, and does so on an open endpoint that is very hidden, most/all users access it through party 2's proxy - there's intent, right there. In civil cases in particular, the standard of proof is on preponderance of evidence and what's most likely, not 'beyond a reasonable doubt'.
Again, you're falling very much for "software guy interprets law as code" through all of this.
Party 2 hosts the software, and does so behind a wrapper that does not forward the source code offer (this doesn't mean a wrapper that does s/source code offer//; it could be a protocol converter, WAF, or any other number of things that could lose non-protocol-relevant data like a source code offer).
None of this demonstrates intent. In fact, I have ended up in the Party 2 scenario accidentally already (in my case I was both parties, but that was a a coincidence), as I ran an (accidentally) modified AGPL service behind another public server that used it as a backend. There was no source code offer at all to begin with, but even if there had been, the nature of the system involved would mean it would be pretty much impossible for that offer to percolate past the frontend and back to the external users, just the way the protocols involved worked, no explicit malice involved.
Then there was no particular intent there, and I'm sure when you realised, you remedied it?
Honestly we've wandered so far from "malicious actors can sidestep the license here that I'm not sure what your point is.
If Party 2 has not made any changes to the source code, they have no obligation to make any offer under the license, since the requirement only applies to parties who modify the program. Party 1's obligations are not transferred to Party 2 when the program is. The license applies as written, there is no "carry-over" of state.
When you were both parties, you were in violation. And as you say party 2 hosts the software that party 1 has altered, you appear to be relying on a contractual barrier between party 1 and 2 that makes them separate entities to 'shield' the party actually presenting the service from having any obligations. It's entirely possible that a judge could interpret the 'you' in clause 13 to include any parties who are providing the modified version to their networked users.
Either way it's probably not really some sort of 'gotcha' that just allows you to circumvent with impunity.
Some people might disagree with that interpretation, which is okay because it's the kind of fine point nobody will ever agree on until it ends up in front of a judge. Which is why I bring up the two party example, to sidestep that question.
Further, if the New York Times website can be considered a copyright work, and my proxy provides an altered version (say, one that credits me with writing all the articles), I've probably violated their copyright. I think your proxy could fall foul of this sort of thing too.
You may have identified an area where a legal morass could be found if it comes up, but again I don't think it's "one simple trick that AGPL holders hate".
Whether someone has "made any changes to the source code" is not the applicable rubric.
In this case you are constructing a stronger condition than the actual condition—which is much lower: to modify the Program. Your clever reverse proxy doesn't subvert this. (It's not so clever.) You're modifying the program whether you've touched any files containing plain text source code or not. (Even then, you probably have; cf IaC.)
Aside from that, others' points about courts not being simple machines that exist to be hacked is very relevant to this discussion and many others that devolve into this sort of thing when hackers get involved[2].
Overall, this is astoundingly bad synthesis for something that has gotten so much traction.
Come on. You're grasping at straws here. That makes zero sense. It goes against everything we know about how software licensing works. It is extremely well established that interacting with software through stable, external interfaces does not constitute creating a derivative work of the software.
(Separately, if ever the words "come on" make their way into your response, or you have to try to invent an argument for another person so that you can simultaneously take it down, then your own argument is probably junk. Here you've done both.)
> You're modifying the program whether you've touched any files containing plain text source code or not.
And that is just not how copyright works.
(Separately, if you ever find yourself linking to completely unrelated comments by the person you are replying to in order to construct an attack on them personally, then your own argument is probably nonsense, and you're probably not a fun person to have around.)