How about a command to alert Sense that a specific type of device is about to be turned on / off ?
During a quiet, stable baseline period the alert might say look for incandescent bulb
which would then be turned on / run for 60 seconds / turned off for 60 seconds
This cycle would be repeated exactly 4 times
This is too simple for complex motor and heat devices or long running time devices,
but it could clean out a lot of the other bubble
I like this idea, albeit much more complicated than it sounds.
Algorithms may need to be changed which could take much longer than its worth?
Although, I completely support the idea. Anything to speed up the learning process is a good idea in my opinion. 
I take it that the signal processing is done in the local Sense box
and I understand the complexity of reprogramming it
How about some mechanism to relay this event back to home base for off unit signal processing ?
You could then send the specific criteria of this waveform back to my sense box ?
As far as I understand the Sense architecture, device detection is done by models loaded into the probe (the local Sense box), but the models are selected, refined and tuned in the cloud, with some human assistance. I’m fairly certain that the models use some form of a recurrent neural network, maybe an LSTM. If you’re familiar with training a neural network, you’ll understand that it’s complex to convert the feedback that “This device is about to turn on/off” into the right backpropagation weights to tune the model for that device.
Here’s an interesting article on the use of RNNs for learning languages, but the end hints at how RNNs learn to understand complex and not completely determinate sequences, similar to power signatures. The article also makes the case for lots and lots of training, before complex structures are “understood”.