Without IFERROR, a common pattern would be IF(ISERROR(A2+B2),0,A2+B2).
With IFERROR, you can just do IFERROR(A2+B2,0)
Excel is such a good tool in many ways, and such a bad tool in many ways. Really experienced power users can follow Excel formulas much easier than blocks of imperative code. But the more complicated it gets, the harder it can be to follow it all. But that also goes the same for software, too. I do think that there is something powerful about the "debugging" you always have turned on in Excel, in that you always know what value a formula has produced, even after it has run. And you can (usually) easily see at what point an error started in your calculations.
For the programmers out there who aren't fans of Excel, or aren't super familiar with it, if you haven't seen "You Suck at Excel with Joel Spolsky" [0] you might be pretty amazed at what you can do with Excel at an intermediate/advanced level.
100%! I like to do that in other cases as well, just to keep formulae simple enough that someone else can easily audit the whole spreadsheet.
I'd rather have 5 extra columns in a calculation, then have a huge formula in a single column. This habit is so strong that I often do the same thing with Pandas: adding extra columns to a dataframe for intermediate calculations, when it would be better to write a larger function and .apply() it all at once.
I've never used Pandas, but I imagine I'd have the exact same instincts to break calculations up as you do, since Excel was my "first programming language" that I was first exposed to in the 4th grade, haha. I certainly didn't learn much advanced stuff at that point, mostly because of the time period and being in a rural Midwest area there weren't a lot of programmers around to learn from and the internet was rather different in the mid 90's. :) But Excel planted the seed of programming in my mind, even though I didn't know what "programming" was.
I commonly use a lookup function in that context (vlookup, xlookup, or index/match).