You are making the same large assumptions that several others seem to be making.
For one thing you are assuming that building a system to cater for heavy load is extremely expensive. It can be, but you really only start to hit expensive solutions once you have exhausted the potential of what you already have. From what you say, you seem to be assuming that NR's system is bursting at the seams because the existing hardware and software is incapable of coping, and therefore more investment is required. Another possibility however is that the hardware and software they have is not the issue (yet) because the basic architecture of the solution is at fault. I have seen at least one appallingly designed web-based solution implemented by a financial institution where they threw money at the hardware and software, but the basic design was always going to result in failure of the service before any single machine even broke a sweat. In that case they didn't need more money to make a better solution, just more thought.
For another thing, you are assuming that massive numbers of NR customers are hitting the web service at the same time. That may be the case, but none of us know that for sure. It could well be that the system is so bad that it flounders even if 20 people access it simultaneously. In any case, any system which is well designed should "fail" gracefully, rather than providing the illusion of working until your connection eventually times out after several minutes.
Ultimately, as with most things, customers will end up with the service they deserve. If people are willing to accept an online service of the standard of NR's, without complaint, then there is no reason for any NR competitors to go to the trouble of building a better system. If enough people deem the recent performance of the NR online service as perfectly reasonable, then we are all doomed to crap online banking for some time to come.