I would argue that what you'd really want to do is lock down the sourcetype of your data, rather than letting Splunk try to figure it out. Most of the time, sending data into Splunk, it will do a decent job of figuring out the shape of your data. If you've got standard log types such as IIS or syslog or Apache access logs, or something like that, Splunk can draw upon its internal library to make sense of your data.
However, as you expand and start indexing larger volumes of data, you'll want to make some efficiency tweaks to tell Splunk how to index the data rather than letting it guess. The proliferation of these numbered sourcetypes is because each time Splunk is looking at your file, it detects a slightly different shape, and so it keeps incrementing that number. I've seen "csv-63" as a sourcetype name.
This is where your expertise as the administrator comes in handy. You know that your IIS logs are writing their output in a certain format, and even the field order. You can provide a sourcetype parameter in your inputs.conf definition to brand the indexed data with that sourcetype. Then, if you need to write your own field extractions, you'll be doing so for just one sourcetype, rather than trying to rename them all to match the "generic" type.
