Hi Erin05,
Not many people have responded to your question because it is very vague. There dozens of reasons why your initial conditions may be unstable. I'm sorry to say the best tool to use to solve unsteady state problems is experience. Unsteady state modeling is as much art as it is technical. If you go about it with a completely technical mindset, you'll have problems.
Some of the more common problem areas are:
Make sure all initial flows at external and internal boundaries match.
Cross-sections that expand or contract too rapidly, both between sections and vertically within a cross-section, may cause problems. Ineffective flow areas may be needed to ease transitions. A rule of thumb is never allow the active flow area of consecutive sections to change by more than 50%.
HEC-RAS methodology was developed to deal with subcritical flow, which was later adapted to super critical flow. Even though you have super critical flow checked, the model always has difficulty dealing with super critical flow at initiation.
You can also mess around with time steps, theta weighting factor, distance of interpolated cross-sections and several other computational parameters that may make your model work, at least temporarily, in order to help identify problem areas.
If all else fails, you can try creating a "hot-start file", which gets fairly complicated and usually can be avoided with a little finesse.
Try rephrasing your question to be a little more specific, you may get a better response.