I think I'll design for 9 or 12 outputs and then add a 1 pps (stretcher) and two TTL outputs, is that enough?
If you've got a PPS output from the GPS already, it's simple enough to just put it through an inverting buffer and drive some lines - even directly from a GPS unit, the adjustment step size is usually <40ns for a lot of modern ones, so you can get good raw accuracy and of course the normal stability over time from just buffering the output directly.
If you want to have TTL squarewave outputs, I would treat it as a mixed domain board, and isolate your analog ground from your digital one. I'd also put a ferrite or a small inductor between the main power rail and the opamp power supply for a little more high frequency rejection, since the digital logic will probably be fast enough switching to make a bit on the power rails even with normal decoupling.
As for whether it's enough.... how many instruments do you really have that can take the input? I've got a scope, SA, RF gen, arb gen, counter..... and 7 outputs is plenty for that, which seems like a ton of equipment already, some of which really don't benefit much from it in most situations... so 9-12 seems like more than enough. That said, it's much cheaper to build in the extras early on, so maybe going all out is nice just because it will be plenty into the future. I prefer BNCs just because the equipment you hook up is usually BNC stuff and the actual 1-2m cables can be gotten cheaply, but decent quality BNCs are definitely the most expensive part of the project if you make it yourself.
I guess one other tip if you end up using a design like the one I linked: if you use a rail to rail opamp and bias them right in the middle instead of the skewed one as I have, you should be able to squeeze a little more amplitude out of the outputs. The way it's setup, driving all four outputs into 50 ohm loads makes the chip raise 10-20C over ambient, so there's not a ton of overhead if you use quad amp chips, but there is a bit available. I just ran into an issue where a low noise quad amp was going to be more expensive if I wanted rail to rail, so instead of getting the most out of my 5V single ended supply, I adjusted the bias point down to compensate for the volt and a half or so you need to be away from the top rail, and that gives you less maximum amplitude to play with.