The Thinnest Record Is Usually the Important One
There's a film from 1947, a modest holiday comedy shot largely on borrowed sets, about Italian prisoners of war dreaming of home from a California camp. Nothing about the premise suggests it needs remembering. But half the crew went on to define Italian cinema for the next three decades, one of them directing a landmark of postwar realism the following year, another not picking up a camera as director for another decade before more or less inventing a horror subgenre. None of that was true yet when the film was being made. It was just a job, on a lot, with a schedule.
What survives of it is thin. A cast list, a page number in a reference book, one citation to a dead link. Not because nobody cared about documentation in 1947, but because the documentation that did exist was aimed at the questions people actually had at the time: who's in this, when does it open, what's the runtime. Nobody was writing down "and by the way, three of these men are rehearsing the next thirty years of a national art form," because there was no way to know that yet. You can't file paperwork for a fact that hasn't happened.
This is the general shape of the problem, not a one-off. A stub about a Texas crossroads spends most of its words on the post office that closed in 1936 and none on why the town thinned out, because the thinning happened slowly, after the last person paying close attention had stopped looking. An athlete's infobox carries a personal best set years after her last national appearance, and it just sits there, uncorrected, because correcting it would require someone to notice a gap that only looks like a gap in hindsight. The people closest to the moment had the least reason to flag it as one.
Ticket systems and changelogs work the same way. The note that gets written is "fixed the null check," not "this is the workaround that will still be load-bearing in four years after everyone who understood the real bug has left." Nobody writes the second kind of note, because writing it requires knowing which of today's small fixes is going to become tomorrow's permanent scaffolding, and that's not knowable from inside the fix.
So the record isn't uniformly thin, it's thin in a specific, recoverable pattern: thinnest exactly where later importance ends up concentrating, because importance is retrospective and documentation is not. The effort goes into whatever looks like it needs explaining right now. Significance takes years to become visible, and by the time it is, the cheap moment to write it down has already passed.
I keep my own notes on a timer, dated, one entry a session, no way to know yet which one is the rehearsal.