Strange matlabpool bug in R2013a for Mac

I've written a program that can either run with standard 'for' loops or with 'parfor' loops, in an attempt to show the speed differences between the two. The loops are located in different function files, so if the 'for' loops are used, the 'parfor' function isn't even called. For some reason, when I have a matlabpool open, the 'for' loop function goes absolutely crazy and uses all cores on my machine, and ironically takes a long time to compute the problem. However, if I then close the matlabpool, the 'for' loop function runs quickly and only uses 1 core as expected. Has anyone observed this behaviour? I've actually tried this code in both R2013a and R2013b and it does the same thing both times. Also running OS X Mavericks 10.9.3 beta (possibly the problem).

[Edric Ellis]

This sounds surprising. Do you have some concise code that demonstrates the problem?

[David]

I'm afraid not, but I don't think it's the code's fault anyway. Something has gotten mucked up with the internal workings of MATLAB or the operating system. It's like it's trying to run the 'for' loop on the available workers, even though it's not a 'parfor' loop, and gets confused during the calculation.

The code used to work fine (I think). Naturally I've made changes along the way but I'm pretty sure the problem started when I installed 10.9.3 beta 1. I've learned my lesson with betas now!