True.

I added “foreseen in the near future” because of thinking along those lines but in all fairness there isn’t really a clear point were the risk of being stuck becomes a such “need” due to “foreseen in the near future problems”.

What I’ve seen done is developing a whole new system in parallel with using the old one as it’s usually significantly easier (and less risky) to reverse engineer the functional and business requirements for the new system from what the old system actually does that it is to get try and put them together from what people think they want, as they are seldom aware of the nitty-gritty details and tend to have only a view of the perfect world usage of the system and not at all of the “what if somebody makes a mistake at this stage?” human error conditions that the system must handle amongst other issues - or in other words they generally “only know what they want when they see it”.

But yeah, that point of mine does relly on quite a vague judgement: it’s better than pursuing what’s new for being new but it’s not a clear actionable rule.