For your situation it depends on a few things. First is how your overall business works. If the server farm is just a small part of the cost of your business, and your product is strong, saving a few bucks on the servers won't matter. But if your business is mostly driven by the server farm and it is a large percentage of your companies expense, you will find out if you are right soon enough..
I'm not sure, actually. There aren't that many physical servers (a couple of hundred - vmware has helped reduced than a lot, already). We are a bit of a "boutique"-ISP in that some of our customers are not price-sensitive at all. Which is good, because we can't compete on price anyway - we compete on flexibility, knowledge, reaction-time. We usually build multi-server, often multi-site, multi-network solutions in heterogeneous environments with high availabilty demands....
Salaries are probably at least as big an expense.
What will or will not happen is that one of your competitors will use an OSS implementation to lower their costs. True, they may have problems in stabilizing it, but they may not. In any case, if they can operate more efficiently that you can, you will have to change. If they cannot operate more efficiently than you can, then you are absolutely correct.
True. Efficiency has always been a concern.
You are saying the same thing many UNIX companies said about Linux many years ago.
Until recently, we mostly used FreeBSD ;-)
Now, a lot of Ubuntu is creeping in...