One of the things that people watching the real time breakdown of twitter but don't ever seem to get is that #institutionalmemory isn't about having enough documentation, comments, test and dev history on the code, database, infrastructure; most places do have that.
Its about the practical real tie knowledge of how those things actually interact in real time.
I am institutional memory for several parts of the programs we test. I was not just the first tester; I sat in on design sessions.
I wrote all the original tests; I trained new staff and oversaw the initial testing. I was the primary tester for years, and despite the fact I only now test this once a year, that once a year is the big one: the full regression. I wrote full unofficial documentation of it for testing with rule and build references. It's still given to new devs to start them off.
The original developers, business analysts, and vendor of these things are gone; I am literally the only one left.
So once a year I'm called in for regression, because though other people have tested it---and I trained almost all of them--no one has all of it. It's not that I actively and proactively remember everything--that's why I keep my own documentation--but because I can tell when something is Eh, That Happens, Wrong Check This, Wrong Okay Get the Devs, or Oh God Nope Call Everyone Now when inevitably, a test fails.
#InstitutionalMemory is remembering the context of how complex systems interact.
Seperis