Because if you know that the latest occurance of somethign will be more likely to be at the end than at the top, it is much more beneficial to start scanning from the back. When you scan from the front to the end on a huge text file such as the 50mb one that the OP said, that takes much longer, because internally each line of the file is tested for a match. Imo though, this feature suggestion is a little too specific for this one feature of reading the logs, I can't see it overly useful in general.
