Agreed on pausing during the vial lifting, if in turbidostat mode, since that also can trigger dilution events. For chemostat it doesn’t really matter much, but it will show up in your data. Spikes can happen for a variety of reasons though, like stirbar randomly jumping or spills so it’s a good question in general.
For smoothing data, I generally recommend doing this in a post-processing step after the data is all collected. Since all the data is saved in text files, you can use whatever software you like to do this. I wrote a set of MATLAB scripts to do my post-processing, but the file names and organization have changed so it needs an update. @mgalardini / @heinsz did you guys ever put up your python post-analysis code?
In general, I set a threshold OD range that I expect data from (say 0-0.8 OD for a turbidostat between 0.2-0.4 OD) and delete datapoints that fall outside that range. This removes the spikes and leaves a gap. I also apply a smoothing function when predicting growth rates or plotting OD over time for figures (reducing down from 10k+ datapoints), but you might choose not to do this.
During the experiment, we use median filters to avoid making dilution decisions based on spikes, but that only works if the spike is shorter than the median filter window.