But wasn't your argument that "Ethereum ... suffers from the unsolvable halting problem" and that "there is no way to prove that a given Ethereum program does not have a bug that will cause it to halt at some point in the future"?
All Ethereum transactions can be proven to halt.
That does of course mean Ethereum isn't really Turing complete, just an approximation.
Any program running on a Turing machine is subject to the halting problem. You may very well find a bug or feature that causes such a program to stop running or halt. This discovery still does not disprove the statement, that it is impossible to prove a program will not halt. Thus your observation does not flow logically.
As pointed out, Ethereum is not infinite, thus not a Turing machine, and therefore does not exhibit a "halting problem".
There's a limit to the number of "cycles" spent on each program in proportion to how much funding they have. It's like a smaller version of the halting problem where you only have to say if a program will halt in five seconds or not.
Turing machines are a theoretical construct, be careful when trying to reason about the real world based on them.
All Ethereum transactions can be proven to halt.
That does of course mean Ethereum isn't really Turing complete, just an approximation.