Problems leading to delays are an inevitable part of any project. As they are inevitable you plan for the unforseen and bear in mind that Murphy and his law can strike at anytime.
However if everything you try and do is fraught with problems you have to ultimately ask yourself if there is not something wrong with your approach, and/or if the project is simply too ambitious for the resources you have at your disposal.
Even when a project can be considered finished (ie machine running according to advertised spec for example) you can always improve matters.
Do you believe that the RiscPC was 'insufficiently tested' because some users reported much greater stability on their machines when the C32 fix had been carried out?