Quite a few modern microcontrollers - most notably a few new PICs and the Cypress PSoC series - have some programmable logic to allow unusual port I/O that previously required some form of external logic. I see that as the start of microcontrollers gaining FPGA-like abilities.
In a nutshell, that is why Intel bought Altera.
What I would like to see is a low cost FPGA with a hard microcontroller inside - it seems like the only options currently available are to use an external microcontroller (more cost/space, limited I/O interconnectivity) or to inefficiently create a soft core in the FPGA and significantly increase the design size.
It's the "low-cost" criterion that is the problem. It might be that the packaging that melds the micro and the FPGA fabric is "expensive," so it's warranted only on the high-end of things (think Zync). Bumping up a size of FPGA to accommodate a soft core might still have you in a chip that's less expensive than one with a hard core processor.
Make that low cost enough and it will be a real hit among advanced microcontroller developers. Now when they have to interface some weird device that's timing sensitive, rather than try to figure out how to (mis)use one of the peripherals to generate the waveforms or bit bang it in software while having to contend with making it work with the interrupts, they just program some logic to do exactly what they want.
I suppose the example here is Cypress PSoC, parts with which I admit that I have no experience.
I've done all of the possible combinations here: separate FPGA and processors talking to each other over some kind of standard bus. I've done PicoBlaze and MicroBlaze and even 8051 soft-cores embedded in the FPGA. I did a design with the Virtex-4 with the on-board PowerPC. I suppose at some point I'll do a Zync design. And I've done stuff which could well have been done in a microcontroller and just did it all in VHDL in the FPGA. What was chosen for each design depended on board space, relative complexity of each part of the design, and of course other external elements. What did the design actually need, what made the most sense? This is the part of engineering that's rarely talked about but is vital for a design to be successful.
Anyway, the FPGA vendors will decide what they will build based on what their biggest customers want, and we get to use whatever they end up making.