Warning message in log

[scott.raynaud]

I posted this question last summer but never received any feedback.  I have a set of about 40 variables, mostly binary, for which I'm imputing missing values.  My setup includes 10 imputations with 10 iterations each.  My understanding is that's 100 passes over each variable.  Sometimes I get the following message in my log for a logistic regression for a couple of binary variables:

 

Warning: Too many iterations, probably due to colinearity with dependent variable

Now, I always thought collinearity was a problem among predictors, but leaving that aside, I'd like to know how this affects my imputations.  Is this really an indication of a separation problem?  It occurs for 1 iteration (1/100) for one variable and 5 times (5/100) for another variables.  Can I safely ignore the warning since the majority of iterations are running?  I don't know what changes I could make other than simply deleting the two variables and re-running the imputation.  Not sure I understand the definition of collinearity with the dependent variable in this context.