What's the basis for this extreme skepticism, is it an accepted law that every problem in programing can be solved with some basic plumbing? My guess for why so many people love the X->Y thing is that it allows them to join in on discussions where they have no idea how to answer the question.
I mean, past that point, you've left "technical problem" land and entered "a description of what the business sees as an exploitable market opportunity."
But to be pithy, it's because the system is all of:
• intentionally brain-damaged for OLAP, by a design focusing on OLTP efficiency to the exclusion of all else (think: bit-packing structs into k-v values in a way opaque to anything other than the program that owns the values; no binary object-wrapper format like ELF to describe which ISA is in use by a bytecode program; etc.)
• very popular (enough that auditors are interested)
• a competitive mutually-untrustworthy multi-tenant platform, where people have every incentive to obfuscate any sort of event logs their programs emit, so that they can keep the competitive advantage of analyzing those logs to themselves
Or, in short: the system is a scrambled egg, and people are willing to pay to have it unscrambled for them. Most developers would just say "you can't unscramble an egg" and give up. But the schema information is there to recover—it's just burned into the logic of bytecode programs. Someone with the right expertise can recover it.
Hi, i need $10.
Why? Because I need to buy a beer.
Why? Because the man at the bike shop is super thirsty.
Why is that important? Because If I sate his thirst he said he’d introduce me to his manager.
Why is that important? Because I want to work at the bike store and it’s the only way to get an intro.
“The traces are the only canonical representation of the state of the system”
Why?