I teach a lab class that is only constrained by some broad learning objectives. I have been using the same project now for 5 years. It now runs very smoothly and I know what difficulties the students are likely to have. I think that is an advantage, but teaching the class is beginning to feel stale and I am worried that hurts the students. Given I could easily chose a brand new project and still deliver the learning objectives, how do I know when it is the right time to change topics.
The interesting question is, does the project feel stale for you or for the students? The students are, after all, only doing the project once, so there is in principle no harm in using the same project year after year.
Of course, there is the possibility that the project that was cutting-edge when you defined it is now an old relic, using methods and technologies that nobody actually cares about anymore. In this case, you should of course move on, but if this is not the case, I see no inherent value in changing the project (especially since any change also means that the first iteration of the new project will be a bit rough around the edges).
My high school Physics teacher changed his course every two years. The reason was that after that time, he would start thinking that what he was teaching was way easier than what it actually was, and he wanted to remind himself of that. Of course, the content was almost the same, but the notes and the preparations were brand new.
A lab is of course very different, and having a rock solid set up is very useful (better than spending an hour just to find out that one of the cables was broken inside, happened to me).
You could take advantage of your knowledge of the instrumentation by using the same machinery, but slightly changing and improving the experiment. For example, instead of reading values from the LCD screen of a multimeter, you could upgrade to a USB one and get the full stream of values (budget allowing). Then, one could try to do a finer analysis of the data. Another option is to look into the residuals of your data from the theoretical curve, and see if there are any patterns beyond noise, and try to figure out why.