There is a known bug in MICRESS 5-31 and earlier versions related to the use of the lower limit for automatic time stepping: If this limit is taking action, the prefactor for the segregation criterion (which is specified as second parameter in the same line) is not taken into account. This under certain circumstances may lead to numerical instabilities!
Bug status: This bug is fixed in beta versions newer than Jan. 2008 and in the next release.
---
original message from Bernd