Removing the randomness from the time to next block will allow us to use the block number as a clock that is synchronized to realworld time. I can see a lot of potential with this!
I think it's impossible to remove randomness completely e.g. the miner of the next block may unexpectedly drop out of the network (as Anon136 mentioned in the interview), or botch the block generation. It'll take time for the network to agree on the next miner in line (on the order of seconds? milliseconds? Should decrease as networking tech and infrastructure improves. EDIT: But increase as the # of forgers increases). So the block number could work as a rough timer if you don't need high precision (e.g. for timestamping creative works), but otherwise I think you'd be better off relying on clocks worldwide, which already have a long history of consensus on time.
In fact, I'm guessing (only guessing, since the great puzzle of TF hasn't been solved) that Nxt will still depend on computer clocks for timing. So might as well go straight to the source for time info.
Whatever the case may be, TF should avoid 15+ min time to next block situations. CfB already hinted at this, the high variance from willy-nilly needle in a haystack random hashing, and using base target to compensate, eventually leads to potentially catastrophic situations like multiple quick blocks followed by an insanely long one.
Q3) In general, am I on the right track with NXTlayers, NXTplugins, crosschain transactions, automated gateways, NXTcash, blockchain FIFO, etc?
I dunno if this is the 'right' track (for BCNext, or everyone else), but I'm definitely interested in all this. The current model of forking to experiment with new coin properties and features is very inefficient.