I don't agree with your timing method. Including the regex generation into the timing is wrong. I the real application, the regex would be set up once at the beginning of the program or even pregenerated and read from a file or even embedded in the source. It can then be reused over and over.

In the application for which this is destined for use, it would be used on each of 500 x 500 char strings, and this process it repeated as fast as the data buffer can be filled with the aim of getting the whole down to less than 1 second, hence the need for speed.

Would you consider timing the repetition loop without the regex setup which is basically a compile time cost not a runtime one? Without the start up costs, I think you'll see a different picture.

As for the limitation on the number of matches, I used 500 as that is the limit of the string length for my application, but again this would be adjusted to need at startup.