i found out that noise~ seems to be seeded only with something like $0+noise_count and not with a better seed like the current time
which leads to the result that if you use noise to control things like in the syn-rain from diy2 the output is always the same and completely reproduceable (well i didn't check if you can even get the same md5)
so i did (a short and dirty) noiser~
i think i overdid it a little bit but at least it seams to give good results
to see what i'm talking about open up the noiser-testing maybe 3 times
for your usual noise~ needs (above 1hz) noise~ should still be enough