The default buffer size is 5000, and that's what I've set my window buffer size at. Since you described yourself as having lots of channel windows, increasing your buffer size by 10k lines has the potential of exposing yourself to storing extra by 10k multiplied by however many of those channels are very active.
The main advantage of larger buffer size is probably being able to use Ctrl-F to find text in the window, since I assume you're not scrolling up and reading them completely. Perhaps a better solution for you is to shrink your window buffer then find a good 3rd party program that can view your channel logs to search instead. (Best not to let it write to the log in competition with mirc.) A 3rd party program might even allow better regex searches like finding text only if the line has <nick1> on it.
Another alternative is to just keep the window buffers small, and if you want to look at one of the channel's history, use loadbuf to put it into a custom window. This is rudimentary, and you can make yourself aliases to let you load entire file, or first lines etc. This loads the last 5000 rows from a file into @log window:
//window -r @log | loadbuf 5000 -r @log $qt(c:\path\filename)
This window won't show the events with the correct color unless you have checked "line colors" in your mirc-options for logging. You'll then be able to search this window with Ctrl-F, and if needed you can copypaste lines back to the original channel.
To use loadbuf or text viewers on a log file, you'll quickly become familiar with how log files combine the channel name + network + date to create the filename, or just use $chan(#channelname).logfile to get the filename for the current log
Thanks for your detailed reply.
I have set the lines to 2500.
I think that would suffice.
