Lord,
I suspect which language you use is much less important than the tools you use to make a professional game. The bar is very high and one person, or even a small team, cannot make all the tools needed. Many of those tools are purchase from 3rd parties for significant sums. At least, that's true for the games that you mentioned. Also, quality of game play is more important that the language it's written in, which is a skill that takes years to learn by practice and failure and lots of testing. For now, Smalltalk does not link to these tools very well. Most are file based which is not so compatible with Smalltalk's image. Managing large art content files and large quantities of them is required which might put a strain on Smalltalk's memory management. Professional games us a lot of tricks at the compiler level and code architecture level to squeeze every bit of optimization at the expense of ease of programming.
Smalltalk's strength is that it's easy to change, but that may be a weakness here. Often in professional games, you don't want the games easy to change by the players because the temptation to cheat is too large.
If you have a game where it's good for the player to change the code, then you may have an advantage with Smalltalk over other games.
If you have a good game design or subject environment niche, Smalltalk might help you make an interesting game one can sell. It allows you to make many design mistakes and correct them quickly. But your niche will be more important than your platform or how professional it is.
That said, Croquet can use Direct X. Maybe you could test out your game design in SecondLife and, when it starts to be popular, migrate it to Croquet.
You might want to check out the Gamasutra game developer site for pointers how to get started for the Independent Game Developer.
Cheers, Darius