I am usually paid for what I develop and open-source...
And after you open source it, somebody works for free and optimize your work and get a few beers in donations. I don't understand your problem.. Miners needs the software to be up do date to earn money. Old kernals are not profitable any more.
I have no problem with you optimizing my work,
as long as an actual optimization takes place which hasn't always been the case lately (I tested what you did on lyra2rev2 side algorithms (blake,bmw,cubehash) and I found absolutely no difference while the numbers you reported were by itself equivalent to half what you would get if you had removed these algos (minus cubehash...), which from my stand point is more or less equivalent as saying that what I release isn't optimized. So at least make sure it is true before saying it... (especially when you say it almost every days... hey I understand that you need to motivate donators, but please don't do it repeatedly on my back

)
ps: I don't deny that you did a good job with quark and x11
Every small change I do might not do any difference, but when you add all the small changes they matter. The changes in blake,bmw,cubehash was so small compared to the lyra2v2 algo that it was almost not messurable, but my implementation use less assembly instructions. After I changed the lyra2v2 algo to use less registers my version is mining at around 4,4MHASH on the 750ti on standard clocks. I can overclock stable to 5,150.
Your version does 3,950. (750ti) (4,6 with overclocking.) My modded kernal is 10% faster...
I do the tests in windows(x86)
You need to merge all the modded kernals together and adjust/increase the intensity to get the speedup...