too much too young - flaws in early internet research

so looking back at some early (not as early say as the arpanet, but early compared to web 2.0) internet research, it looks to me like it is really worth re-vidsting some of the "well known" results from between (say) 1989 and 1999 (i.e. the decade before the last 10 years or so).

I can see several "well known" results (a.k.a. folk knowledge/science) that are probably wrong - often, these are architectural in scope, and are because someone (often from a well known institution) wrote a preliminary paper on a prototype (lets not mention names, but early RSVP code or IPv6 spring to mind) and drew conclusions which were taken, for want of a better word, as seminal - so some comments on scalability (e.g. of intserv and rsvp) and security (e.g. of 8+8), put people off working further in promising directions, and were not at all correct - we don't know what the right answer was because no-one has re-done the work in the context of later knowledge e.g of switch router design, or of crypto-assigned addresses, or of hardware support for fancy queueing and scheduling - there are lots of examples - the problem historically (and for the community and services and products) is that with an exponentially growing network, the tipping point is past, at least as far as today's IPv4 internet is concerned. But I worry that newcomers building new systems (e.g. IPv6 internet in china with Huawei) may take the old research as correct when it isnt.