Science is not totally dependent on repeatable experiments. That is only one method science uses to test data.
The fossil record is supported by experiments in dating etc., but there are few experiments you can do directly. The main factor in eliciting evidence for evolution from the fossil is forensics. Forensics are backed up by experiments. These support the analysis of the fossil data.
When different disciplines all correlate a conclusion, no experiment is necessary.
And how is it that you are able, from this information, to rule out the presence of an intelligent agent?
Consider again the examples of distinctive dog breeds, or gentically modified corn, which we know to have come about due to the actions of intelligent agents. Suppose (again) that the a priori knowledge of intelligent agents is somehow masked from the body of evidence available to a scientist. Would "forensics backed up by experiments" be able to identify the presence of intelligent agents in these processes? Would a scientist even bother to look for such a thing?
I believe the answer to the latter question is "no," as the going-in assumption is that changes within and between species arise from natural selection.
The scientist would most likely form a hypothesis about which natural selection pressures might have led to the particular traits, and that would be that. A plausible hypothesis, supported by evidence, and yet incorrect because the real cause was something other than "natural selection." This is why I'm suggesting that the theory of evolution, as usually stated, is incomplete.
Which brings us back to the first question. We know that, in some cases (such as the two above), intelligent agents have been active in the development of species. What sort of test could show their presence? That's a tough one, and yet we know that the demonstrable incompleteness of the theory of evolution would seem to demand tests of that nature.
It appears to me that the argument about ID basically revolves about this point. Current tests may not reveal the action of a hypothesized intelligent agent. The ID side may claim a hypothesis of intelligent agents, but the evolution side invariably dismisses it because there is no evidence. And the reason there's no evidence, is because there's no test that the evolution side recognizes as valid.
A truly scientific resolution of the issue would require the development of some agreed-upon criteria for recognizing the signature of an intelligent designer in the loop, or at least for looking more closely into the idea.
One of the characteristics of this argument is something to which you have alluded on this thread, and which some others have stated more or less explicitly: they claim that "evolution is a fact," and therefore see no need to formulate tests for alternate hypotheses. This is a great way to close off debate, but as we have already seen, it leaves one with an incomplete theory.