This is actually quite a good question which caused me to scratch my head a lot. There are several issues at work here, and it is not always clear which one dominates the overall outcome.
First, gold is not so inert as you might think. If you take gold powder of flakes, and drop them into a cyanide solution through which you bubble air (hardly extreme conditions), the gold dissolves to form [Au(CN)2]- ions. Gold is oxidised by plain air at elevated temperatures (100 degrees C) but the oxide Au2O3 is not particularly stable and decomposes back to solid Au and O2 at slightly higher temperatures. You can prepare Au2Cl6 by heating gold in the presence of chlorine gas at about 200 degrees C. From this compound you can go on to organometallic compounds which are quite useful. So while a little stubborn, you can get it to react.
Now as to the mystery of Aus lone 6s1 valence electron. You would expect: higher quantum number, more distance, less attraction, easier gotten rid of. But that's not how it works. You have to work out a couple things here. One is that orbitals become bigger as their principal quantum number increases. It simply puts them further away from the nucleus, but not in a linear fashion. Second, electrons repel each other, so they don't want to be in each other's neighbourhood. Third, electrons tend to 'shield' one another from the attractive force of the positive nucleus. Electrons which are in higher orbitals therefore experience 'less' positive charge. Where electrons end up is of course where all these effects are balancing each other. If you do the math, you will end up with the rather surprising result that despite its high principal quantum number, the 6s1 electron of Au is more tightly bound to the nucleus than Ca's two 4s electrons! In other words, the increase in Z is not offset sufficiently by higher shielding and higher quantum number to make it experience less attraction.
And just once you think you understand it, in steps another effect, namely that of orbital overlap. In order for atoms to bond properly, their orbitals have to overlap sufficiently for a bond to appear. If you take Au+ (so without its 6s1 electron), it can only offer 5d electrons to anything whishing to cuddle up. (Well, perhaps a little 5p too, it depends.) Atomic oxygen can only offer 2p orbitals, and the overlap between a 2p and 5d orbital is not good. That is why in Au2O3, there are [O2]- entities rather than O2- ones. [O2]- is simply bigger and can thus provide better overlap. Chlorine can offer 3p electrons, still not perfect, but better than 2p ones. And so forth.
You can go a step further and look at the particular shape of the 5d-orbitals available for bonding, as some will not overlap, and others will. If you take out electrons from the 5d orbital, some of the remaining ones will slightly shift position, becoming more elongated or contracted to accomodate for the 'gap' (and thus change in repulsion). This complicates overlap calculations (and thus which compounds will form) considerably.
In effect, what you learned that orbitals must be filled for ions to be stable is only an approximation to the real thing. For example, you can let the noble gas Xe react with PtF6, or even pure F2 under relatively mild conditions, even though the simple approach predicts that the filled valence shell of Xe is immune to reactions. It is not: apparently the system can exist in a lower energy state by overlapping and mixing some of their orbitals, and is thus susceptible to chemical bonding.
Finally, the terms 'ionic' and 'covalent' bonds are just descriptions of the extremes of orbital overlap. Sometimes the shape of the resulting bonding orbital is centered predominantly on one atom. That is an 'ionic' bond, because effectively, one atom becomes negatively charged, and one positively. Sometimes it is centered right between two atoms. That is a covalent one: neither atom has a particular charge.
Now I have to cross my fingers and hope a real chemist doesn't frown too much at what I wrote... :-P
