I noticed the temperature gauge reading higher than usual on my way home Thursday. I made it home okay, and after it cooled down I went out to check it. It was about a gallon low on coolant. I filled it back up and started it to see where it was leaking. It wasn't obvious at first, then I noticed a stray drop or two flinging off the belt. I couldn't see a wet spot yet, but I could tell it was leaking onto the belt somewhere. The warmer it got the more it leaked. The belt was getting wetter, and stickier from the antifreeze, and it was starting to make those "I'm going to throw this belt off" noises. And I finally saw it as it started really gushing from behind the water pump pulley, so I shut it off. It never got too hot, only about halfway up the gauge. Mine usually rides just barely above the quarter line on the gauge.
So anyway, all that background information aside, I went to get a new water pump before I took the old one off. After I got the old one off I could see a definite track where the coolant has been coming out of the weep hole at the bottom of the housing; otherwise the bearing on the old one feels tight, doesn't wobble at all, and there is no shaft play. It spins just like the new one. The coolant track had a rusty color to it.
On any other water pump I have seen fail, it stayed pretty well sealed, until the hub started wobbling and it was obvious there was a problem with the pump itself. Which brings me to my questions.
1. Is this how these water pumps fail? I'd hate to get everything back together and then find out I should have done something else to clean out the block to save this new pump. The mosquitoes got to me tonight, or I would still be at it. I'm thinking about going back out and squirting it with a water hose to clean off all the antifreeze stickiness under the hood. I'll take the thermostat out and try to flush out the block while I am at it.
2. Am I supposed to use sealant on this water pump gasket? I have heard some conflicting reports about the use of silicon or RTV in fuel injected cars. What I remember hearing is that there is a chemical in the RTV that has an adverse effect on some of the sensors used on modern cars. I have had fantastically good service out of all the fuel injected vehicles I have had, and I have never had occasion to use silicon under the hood of any of my previous fuel injected cars, so I have no direct experience with this issue one way or another. Just to be safe, I got a tube of silver Permatex RTV labeled "sensor safe," just in case I need it.
I looked on the factory FSM (thanks K) and it didn't say anything about using sealant. The block surface where the gasket will seat looks good, although that is what I was working on when the mosquitoes attacked and didn't get very far. In a perfect world, the block and the water pump gasket surfaces should theoretically be on a perfectly flat plane, and the gasket should be all that is needed to seal it. But the world I live in is far from perfect, and it seems like I guess wrong about half the time. I'd like to get this right first try, so any information will be greatly appreciated.

pumps fail for a number of reasons.
Sometimes the rubber seal just fails while the bearings are still good. Othertimes the failed bearing takes out the seal.
I allways run the pump sealing surface over with a flat file, then apply sensor safe permatex gasket sealer to the gasket on both sides. You could use RTV. Never had a pump leak and I like it that way.

Is there a trick to tightening the pulley bolts?
I got it all back together. I am going to wait until Sunday to fill it and check for leaks. I want to be sure to give the sealant a lot of time to cure. I think a healthy person could change one of these water pumps in a couple of hours without much trouble, if everything goes smoothly like mine did. I have to stop and rest a lot, so it takes me a long time to do anything. It gives me more time to plot and scheme, though.
A couple of things I didn't see mentioned anywhere else on site, that are somewhat common sense type things, but you don't want to leave out.
Clean all the bolts. I used a hand held brush shaped like a toothbrush, with brass bristles. I was wishing I had a bench grinder with a wire wheel before I finished. A media blast cabinet would come in handy here instead, just make sure to degrease the bolts so you don't contaminate the media.
Put sealant on the bolt threads. Some of the bolts holes are open to the water jackets, without sealant they may leak coolant. Also the sealant may help reduce corrosion on the exposed part of the bolt. If a little bit squirts out the hole when you tighten it up, that'* okay too. It helps seal the bolt head.

I'm not sure I got the pulley bolts tight enough, I couldn't fit my torque wrench in there. Other than that, this is a pretty straightforward swap. I would have rather done it a little later, when I get my Grand Prix back, but sometimes I have to eat the plate I'm fed.

Yep. The seal on the impeller shaft goes bye-bye and it starts to leak coolant out that little hole. (which is why that little hole is there in the first place..it'* on nearly every water pump made...to drain out anything that gets past the first seal on the shaft)

Now, once you get it all back together with a new water pump, check your belt tension. Too tight is just as bad as too loose...too tight puts a lot of stress on the bearings that support the pulleys and the water pump gets it hard being a 180 degree change in direction for the belt.

That is reassuring. Most water pump failures I have seen have been bearing failures that made noise but didn't leak externally.
I am glad I made it home okay and didn't have to change it on the side of the road. My car shows 98,000 miles. If I still have it in another 100,000 miles I will go ahead and change the water pump before it starts leaking. As easy as it is, it is a good peace of mind investment.
I got it filled Sunday evening and drove it around a little. No leaks, no noises that aren't supposed to be there, and no overheating. Now I'm back to showing everybody my wrong rear bumper.