nprofile1qyt8wumn8ghj7un9d3shjtnyd968gmewwp6kytcqyqpjn4vqgkchzj5qlpg7c0h2p8404a6memc63zcwdhyanvmx3pdauyvsyqt (nprofile…syqt) Preaching to the choir here, no doubt... but... my mind goes:
Can you run an FFT on the 38us bursts to see if they have any energy at a particular frequency, possibly pointing to a source clock domain that you might recognize?
Anything running at 2.3kHz period at that low a duty cycle is most likely of software origin as nothing of hardware origin would have such a long gap.