Does anyone knowledgeable know whether it’s possible to drastically reduce the overhead of reading from procfs? IIUC everything in it is in-memory, so there’s no real reason reading some data should take the order of 10us.