Cool, No worries.
Um, none. Use RPVST+ unless you are working with a multivendor environment. And If possible, just don't use any STP at all. I guess you should still configure it, stop stupid people from breaking stuff, but a properly tuned routing protocol will offer faster convergence and more predictiable operation so if you can taker Layer 3 to the network edge, do that. Otherwise there is a great range of products that offer support for Multi Chassis Ethernet through switch stacking and VSS, etc. If you do have to use some form of STP, try and keep it to the edge of the network in small domains (ie Within 1 network closet or at worst within a building), use RPVST+ and spend some time tuning it (timers) to achieve faster convergence.
As I mentioned earlier, any per-VLAN STP works well in a case that you have redundant paths back to the L2-L3 interface It's an old network design I'm slowly making go away now but we used to have approx 150 VLAN's spread across 2 core routers running HSRP on both. Adding RPVST+ to this and configuring properly worked perfectly. The systems had different priority for each HSRP instance making one router the prefered active unit, load balancd by prefering different routers be active for different VLAN's. We used RPVST+ and set the prefered active router as the root bridge for that particular VLAN and had all our distribution layer switches dual-connected to each router. For every VLAN the link to the standby HSRP router would block and the link to the active HSRP router would enter the forwarding state. It didn't happen a great deal but you could then also run dual uplinks between distribution and access layers (one access switch to two distribution layer switches), since the HSRP was evenly balanced half the VLAN's would use one upline and half would use the other and enter the blocking state for opposite VLAN's. No tracking of interfaces and changing HSRP priorities etc. If a core router failed HSRP became active for everything on one router, if a distribution layer switch lost access (contractors dig up a fibre) the cross connect between the core switches/routers would enter forwarding state for those VLANs and continue to provide access. You get the idea, same thing again for access layer loosing one connection to distribution layer. The problem of course is that the network is so flat it becomes a pain to manage. Hence it is now going away for a much better model that invloves less layer 2 spanning and more layer 3.
