One potential options is to use a Triscend E5 device. While the E5 is not quite as fast as a Cygnal device at the same frequency, its on-chip programmable logic operates significantly faster. Triscend calls the embedded programmable logic "Configurable System Logic" or CSL for short.

For example, you could have the nine 16-bit counters implemented using approximately 144 CSL cells on an E5 device. This would put you in the smallest and most cost-effective E5 device called the E502.

When the switch is closed, this could enable the associated counter. When the switch is opened, the counter stops counting. With some additional logic, each counter could flag an interrupt to CPU indicating that the switch is open again. You can also set it up so that once you read the counter, the counter is reset.

With the E502-25 device, each counter would operate up to about 35 MHz. Using the faster -40 device, each counter would operate up to about 50 MHz. If you wanted even finer resolution, it is possible but costs a few more CSL cells.

The counters are completely independent of the CPU. The only CPU loading would be from the CPU servicing the counter interrupts.

It you are interested, there is more information about the Triscend E5 available from the following link.
http://www.triscend.com/products/IndexE5.html