It could be that Volta is for commerical use and Pascal for consumer use?
For 2017 I agree, pascal dies will cover the consumer/pro market 2016 and 2017 (maybe volta Tesla products released after contracts filled whenever that may be).
Looking at the Sierra and Summit destined Volta chip I was wondering just how far they go in dedicating space to features that are of no use to the consumer that it will not be suitable to ever package/market it as a consumer card. (Not because it couldn't perform as a consumer chip but that another die could achieve the performance or more at a smaller die size and sell for better margins as a high end graphics card - no need for masses of fp64, NVlink support, perhaps fewer HBM controllers and without any other transistor eating tuning for HPC).
Once these commercial orientated features are established and expected in follow up commercial products I can see that pattern recurring, effectively a commercial only die.
It does depend on both commercial and consumer markets being large enough/exploitable for long enough before the next process shrink to make it worthwhile, and with sufficient performance progression to make it marketable (GM200 showed plenty). Otherwise you may as well throw supercomputer Volta into the consumer market for a few months and await 10nm. /Speculation