Actually any organisation worthy of the label "engineer" should do a thorough, no-blame, and collaborative post-mortem of such events so they don't happen again, and that would include warnings given.
Feature on time + risk, feature late + no risk (and anything in between), it's in the end an engineering and business decision and either choice might be the right one depending on the circumstances.
Absolutely, this is a often ignored or missed aspect of building systems. Quantifying risk in a fashion that all parties have the same understanding of said risk can be very difficult. There has to be a common language and agreed upon thresholds of risk. After developing shared understanding of risk, it's much easier to highlight and change the behaviors of individuals and the teams.
