Looking for some guidance on Hackers, Or aren't we all? Options

[Professional Lur...]

I have recently picked up SR4 and was very interested in how hacking was now handled. Since in every other SR game I've been in the GM encouraged the players to let him NPC the decker, I was curious if this edition made it easy, quick, and fun to have a PC hacker.

I plan on running a one shot with a hacker to test it out, but wanted to know how well it has worked out for everyone else here. I have read many of the threads on the issues, and read some house rules, but haven't found solutions to some problems that are still bothering me. Until Unwired is released, you guys seem my best bet on getting some good mechanics in the meantime.

I guess I really should start by asking, do we really need to make changes? Sure, there are some fuzzy areas in the rules, but do we really need house rules or just clarifications?

For instance, if someone uses Exploit to get in as admin, does he simply pwn the system (turning off IC and alarms, etc...)? At first this sounds too powerful, since it's just an extended test they will eventually win, but would it actually be that bad in play?

In the security section of the book it talks about some networks not being accessible from the outside (anti-WiFi paint and such), forcing the runners to drag their hacker along with them to get at the important bits. At that point, the challenge comes from physical and magical security, plus the possibility of matrix alarm, since the hacker would be using the harsher "hack on the fly" rules. Is that enough to prevent admin rights abuse?

I have some ideas on how to make the hacking for admin harder:
When the node gets to roll firewall + analyze, add +2 if the hacker wants security access, or +4 for admin.
If that's not enough, have the system roll analyze alone w/ +2 or 4 for level of access every time the hacker does something on the system (or once a round). Maybe only do that on nodes that require those passkey things the book mentions once in fluff but never in rules. Give the team a reason to try and steal one, or put up with more risk online.

I'm not completely convinced these are needed. I want to try the game first before adding house rules, you know? Until I do, I would really like any of your opinions.

There are other questions I have:
Is a node a network? If I hack into a business do I have access to data on all their computers, or will I have to hack again to get into the CEOs? Would I just send that other machine a command to give me some data?
What about agents and IC? Could I run more then one off my comlink (or loaded on the hacked node), and have them attack any IC I run into? If I can do that on my comlink, why shouldn't IC do that on it's node? Is Response lowering the only reason not to do this?
What about multiple comlinks? Can you hack with two response 5 comlinks, one with you programs and the other with a number of agents loaded? Or multiple Nodes on a network with IC loaded and attacking you on the node you got in?

I think the Agents/IC is actually the only thing I _really_ have problems with. Lots of agents/IC means lots of dice rolling. I think that can be fixed by having each side use the assisting and auto success rules (agent 4 + attack 4 = 2 extra dice to the hackers attack), but that still doesn't seem right when the hacker has 3 or 4 comlinks worth of agents attack 3 or 4 nodes worth of IC. That way leads to escalating madness. I need some kind of cap on how many agents/IC can be involved, and not just response since more then one node can be involved (right?). Maybe give the systems analyze roll (to get in or with the house rule above) +2 dice for each agent you have loaded (so as to limit the agents loaded before the alarm)? Make it take a roll to get an agent loaded in alert mode(to limit the number once the system is alerted)? Make agents take a complex action longer to load for each program they have loaded in themselves (not something you want to do while being attacked/traced by IC)? And that only limits the hackers side of things, what about the IC?

I'm sorry this has ran long, but I could really use some help figuring some of this out, particularly the agents/IC problem. Any comments or suggestions? Or threads I missed?