I don't know why something as expensive as the SE-3 doesn't have two loop bandwidths, one for the unlocked condition, so the unit will lock more quickly and over a wider range. The second loop bandwidth would be very tight and would only kick in when the loop is locked, thus obtaining the improved noise rejection and fidelity you speak of.
As for the lock up time and switching between carriers in a break-in QSO, this is easily overcome by defaulting to a diode detector until the sync detector locks. Once locked, the output is soft switched to the sync detector. It is totally transparent to the user and you would hear no whoops, swishes or heterodynes. Doing this in software should be possible too.
I had a chance to play with Chuck's SE3 and it worked very well. When it is done in software you get the loop performance the programmer designed in so there is no selection of loop filter constants like the SE3. The tighter you make the loop the better it sounds...but the harder it gets to acquire a lock. I've had a number of emails with Alberto the SDR master on this subject. I think he has improved his software. Flex software has a much wider capture range so it doesn't have the fidelity of the SE3 when the tight loop is selected. This is all just a software issue and with the right software it would be possible to duplicate the performance.
Alberto's VER .99 sounds the best when locked and has a nice phase meter but has problems when you listen to a round table and everyone is on a different frequency forcing you to reacquire the lock each time someone transmits. The SE3 would do the same thing if the loop switch was left in the tight loop position.
Chuck modified the loop filter if I remember to make it sound better.
I bet Sherwoods next SE is a DSP.