Of course, I now think that many pressure spikes are now real events (e.g. dust devils). However, I've seen other spikes (mostly upwards) that aren't real. Further tracking finds that most are (only?) spikes in the Paro's temperature measurement. These spikes are always positive. During 10 Aug, most of these were with t.8b, however I also see them on 5b, 9b, 10b, 7t, and 8t. These are not exclusively on pressure2, nor are they all the channels on pressure2. Also, there is a pattern of most (but not all) spikes at night. I <think> these started in the last, narrow, configuration, but I'm not yet sure.
Since Paro t changes slowly (essentially a Tcase), it is trivial to despike these. However, despiking would best be applied upstream of Gordon's code to apply calibrations, and I think it would be a lot of work to redo this at the moment. (But it should be done before preparing the final data set.)