Particle.publish is blocking for seconds at a time in multi threaded mode

[jaza_tom]

Update:

I posted a related issue here:

Cellular.getDataUsage is blocking application code for seconds at a time in multi threaded mode

Update: I have found that blocking issues go down substantially if the modem’s data registers are queried at much faster frequency. I modified above code to wait only 500 ms in between calls to Cellular.getDataUsage() and now it is only blocking for about 20 ms per call. If the network connection is lost, then the first call to Cellular.getDataUsage() after the connection is re-established seems to work. Calling Cellular.getDataUsage() which Particle.connected() is FALSE seems to screw up th…

It seems that if you keep hammering the modem with Cellular.getDataUsage() (on the order of twice per second) once an Electron has successfully connected to the internet, then subsequent calls to Particle.publish() don't block for very long.

My working theory is that if the modem doesn't receive an AT command from the STM32 regularly (say for more than a second) then it goes into some sort of lower power quasi-sleep mode which makes subsequent AT commands take longer to respond to. So hammering it keeps the modem "awake", thus reducing latency associated with synchronous AT commands.