First, I noticed that of the five long across answers, four had years in their clues and one didn't --- and since it was for a city, it would presumably have been easy for Matt to work in a year if he'd wanted. ("Founded in....") So I deduced that the absence of a year was significant, perhaps to indicate that this entry was special in some way.
Perhaps it told us how to interpret the other four. I see "gr AND f OR ks" but that doesn't suggest anything.
Or perhaps it wasn't actually a real theme entry but was only there because symmetry required it; this made sense in part because we were looking for an eight-letter answer and that's easy to get out of four answers.
I tried using the non-zero last two digits of the years as indices into the answers and into the clues. There were eight of those, right? But no dice.
OK, I'm clearly overthinking this. What else can I do?
Well, the title is suggestive. "Nine Down Is Like Five Across". 5-A is "E-Bay", which is unusual in that the first letter represents itself, not part of a larger syllabic unit. How many down entries are like that? Nine of them! (iPods, TCBY, etc.) So there's a set of NINE DOWN entries that are LIKE FIVE-ACROSS. What do I get if I take those letters? TXATMAIGR. Hmmm... what about if I take the first letters of those clues? FAIMMNSMS. Perhaps those can be anagrammed or Caesar-shifted? Nope. Well, maybe I'm missing something, and they decode to AUSTRALIA and the answer is CANBERRA? If this were week five, I might go for such a leap of faith, but it's only week three, and everyone who's commented publicly says that once you have the "a-ha!" moment it's a quick and obvious solve. I don't have an elegant answer, I have a kludgy guess.
Maybe I'm on to the right idea on how to parse the title, but I have it backwards. After all, the IS is pointing the wrong direction for my previous hypothesis. Maybe there is a pattern in the long answers that can be used to extract capitals from them, and the same pattern can be applied to SCATHE, thus NINE DOWN IS LIKE FIVE of the ACROSS answers. I can insert a letter into PARS to get PARIS and into TOYO to get TOKYO, but I can't get any capitals out of any other long answers, adding one letter to SCATHE would only get me to 7, and there are no world capitals with the letters SCATHE in that order.
OK, I'm REALLY overthinking this.
Maybe Matt's being devious and "world capital" doesn't mean a city, it means a currency. We have YEN in the puzzle.... but I'm really not getting anywhere. I can't even name an eight-letter world currency.
It's supposed to be easy once you see it, right. Let's go back to basics. What do the theme answers have in common?
I see SENSE and HEAR but nothing else useful. PARSE, maybe?
Finally, nudged by Joon's tweet about depending on Google, I looked up Grand Forks. Imagine my delight when it said "Grand Forks, ND" instead of Michigan, where I had mistakenly placed it in my mental map. And of course the ND was exactly what I needed, because then I realized that the preceding clue entry had studiously avoided naming NANCY DREW. So I Googled the movie line and saw NAPOLEON DYNAMITE, and assumed that the coach was from NOTRE DAME and the song was by NEIL DIAMOND. (Perhaps if I knew something about pop music, sports, or movies I'd have had an easier time with this puzzle.)
Lesson 1: Before overthinking, Google all the partial answers and see if any similarites pop out.
Lesson 2: When a piece of information is not really and truly well-known (like what state a city is in), double-check it.
Interestingly, for the last few weeks my task at work has involved trouble-shooting various web problems. I've been staring at large volumes of logs, and writing scripts to summarize and graph the data, to correlate different types of logs from different systems, to look for anomalies and research what could be causing them. This has a lot in common with solving metas --- there's a lot of information, much of which looks promising but turns out to be irrelevant, and some of which can't be properly interpreted until I double-check my assumptions and do some web-based research. And in both cases, I'm certain that there IS an answer. In the case of the meta, it's because Matt put it there; in the case of the work stuff, it's because we didn't intend for the system to work this way so there must be a cause for the discrepency between theory and practice. Solving metas is really good practice for how to think about this kind of problem-solving, and has applicability to the real world and my paying job.
The lessons I took away from Matt's meta are not just lessons for solving meta's. Neither is the list of "Have You Tried..." which my Mystery Hunt team uses. OK, interpreting web server response times as Morse Code isn't likely to be useful, but filtering to focus only on the top decile of values may be (hey, look, they all have the same value for the referrer field; I didn't expect THAT!)