Re: The bottle neck is further up the chain
"1) no one sits there and hammers their port at 1Gb/s all day long"
It doesn't matter. If you have network /home and fileservers then you need to cope with all the machines starting up first thing in the morning and after lunch WITHOUT slowing down.
That means designing the network to cope with worst case scenarios, not the usual 1-2% utilisation.
"2) the switch uplinks are the contention point for sites with local servers. who has more than a pair of 1 Gb uplinks from their access switches?"
When access switches with 20GB/s uplinks are between £300 (basic) and £1200 (fully managed), with the core switches at around £5-8k, why would you NOT run 10GB/s uplinks? (Hint: my stacked access switches run 20GB/uplink per 48-port shelf in the stack - this also copes with $IDIOT unplugging cables in the switchroom - it takes a lot of disconnections to knock out any given access switch AND with TRILL+distributed L3 in the core it takes a lot of disconnections to knock out the core network. TRILL isn't just for data centres and was never intended as a datacentre protocol)
"3) if your core servers are plumbed in at 10Gb/s and you have 100 users at 1Gb/s where is the contention now?"
Why are you only running a single 10Gb/s connection on a critical server?
For that matter: Why do you only have a single critical server? Where's your failover and redundancy?
"4) your 1Gb/s wired connection is pointless if you've got less than (total users x 1Gb/s) WAN link if all your servers are off site."
This is where you explain to the PHB that running a local server room costs £N and running 10GB/s uplinks costs 5*£N - with installation of the redundant physical 10Gb/s path having a 20*£n installation charge - and don't try to shortcut this as it costs £30k/hour in staff salaries _alone_ when the systems are unreachable.
"5) the bottle neck is either the switch uplink, server uplink or WAN link, fast wifi with lots of users on does not change this"
100 dual-radio 802.11ac WAPs (one per office - effectively one per 3-8 wifi devices) and a pair of suitable WACs will cost you between £50 and 80k depending who you buy from. That factors in 20 outdoor, reception area and meeting room units to handle the more complex cases.
The trick is to design _properly_ and make sure that bottlenecks are eliminated in the first instance or accounted for in such a way that you can slot in upgrades to cope. That cheap installation isn't so cheap when you have to toss the whole bloody thing to install something which works properly (and can be incrementally upgraded)