Why is a MARC record structured as it is?

Part of it is oldschool programming language protocols (space was precious thus the wealth of abbreviations in earlier records) and part of it - at least what is taught is - based upon the card catalog model. Many of those fields also have field limits in OCLC. 

In terms of the order - it tried to put the "most important" information first. The stuff for the computer to display properly and read the record (the fixed field stuff) and then Author, title, physical description. The only reason I can think for putting subject headings last is that there may be so many it would put the author/title too far down (following old school card catalog style as well - they might end up on a separate card). This is a good overview http://www.loc.gov/marc/umb/um01to06.html.

Also, many libraries "produced" print card catalogs from the MARC record for many years after conversion. MARC records are heavily influenced by the card catalog. If you want to see the full range of MARC (with ISBD punctuation) see oclc.org/bibformats/ ;-) I've read that only about 10% of MARC fields are actually used, but that is probably dependent upon the type of materials and library.

The problem with MARC is that it doesn't work well in more modern programming systems. When we start to look at E-R / relational databases that break data into tables and then into fields -- well, MARC doesn't really do that.

Instead it's one long string of data only divided up by field tags and subfields vs. structured tables with fields in separate tables. I'll write more on this later.