Once equilibrium is reached, we can expect hashing rates to bottom out. A similar rush will occur over time as the cadence of Bitcoin hashing technologies approaches more regularity, not unlike how new generations of CPUs and GPUs are released on a timely and predictable basis.

This will be the pattern of Bitcoin hashing -- waxing and waning with each new generation of ASICs and the lull between each release.

One has to wonder what will happen to all the surplus ASIC hardware as well-financed miners regularly upgrade their hashing infrastructure.

Short answer - it isn't likely to change much.

In order to significantly affect the overall bitcoin issuing rate long-term, the hash rate would have to continue to grow exponentially for an extended amount of time - years. In reality, the hash rate can only grow exponentially while an older mining technology is being supplanted by a newer one. It happened already with CPU to GPU/FPGA, and now it's happening again with GPU/FPGA to ASIC. It will grow exponentially while everyone hashing upgrades to ASICs, but once the hashing market is saturated with them, it will drop back to modest linear growth, as the only growth will be either new miners coming online or modest, linear improvements in ASIC technology.

Haven't looked at this part, but I have a feeling that the difficulty itself wouldn't be able to go high enough to support long-term exponential growth in hashrate.

It's not an exact science since no one knows future network increases, but you can see a detailed view of the thought process around that question here:

http://thegenesisblock.com/at-this-rate-the-last-new-btc-wil...