If we consider Treatment (Method) values in columns and samples in rows, to use Friedman Test to analyze the significance of difference, how big can the matrix be? What are the limitations for the number of samples and treatments?
In theory, the Friedman test can work with a data matrix of any size.
Computationally, SPSS cannot accept more than 2 billion rows (http://www-01.ibm.com/support/docview.wss?uid=swg21476061).
In practice, you never want to perform a formal frequentist test with huge data: there will (nearly) always be some tiny effect size and, due to the huge sample size, the p-value will be practically equivalent to zero.
I have not come across any literature indicating that the Friedman test loses (gains) its validity with varying sample sizes. Some commentators have suggested that an alternate approach would be to extend the more powerful Wilcoxon sighed rank test procedure, because it utilizes between block comparisons.