Colorado DWR has a great document that discusses what happens (page 19). Below is a link to the document and the text that discusses it.
https://dnrweblink.state.co.us/dwr/ElectronicFile.aspx?docid=3566962&dbid=0
HEC-RAS has the capability to model the pipe with an initial piping elevation set at the bottom of the dam (most piping failure situations); the piping hole is modeled as a rectangular hole, which is more consistent with the final trapezoidal shaped breach section, thereby reducing discontinuity. The bottom width of the hole enlarges proportionally to the final bottom width according to the selected progression, as does the height of the hole toward the final breach depth. This will make the hole height/width ratio greater than one if the final breach parameters chosen show a bottom width narrower than the dam height, but since the orifice flow is based upon the area of the orifice and not the width versus the height, this is a valid assumption.
Once the water level drains down to the top of the enlarging piping hole, the crest is assumed to collapse and the algorithm transitions to weir flow. The bottom width and the top width of the breach continue to enlarge laterally until the final defined width and side slopes are reached. Unlike HEC-HMS, HEC-RAS never assumes orifice flow with a reservoir level below the top of the piping hole. It should be noted that in several modeled case studies, the crest collapses when the height of the piping hole reaches near 0.6Hb and the peak flow occurred in the weir flow portion of the failure soon after the crest collapses. This is dependent on the drawdown rate versus the breach progression, and selected final parameters, but helps support the reasonableness of the program.