For further technology scene context in 2007-2009,
"Grid computing" links to a number of distributed computing projects: https://en.wikipedia.org/wiki/Grid_computing#History
IIRC, there was a production metric-priced grid system developed around Seattle/Vancouver called "Gold" (?) that was built on Xen and is likely a precursor to metric-priced Cloud services like EC2 and S3 (which now simplify calculations for how much a 51% attack against a Proof of Work txpool with adaptive tx fees costs with n good participants in the game) which incentivize efficiency by penalizing expensive operations.
Code bloat was already a thing: how is everything getting slower when Moore's Law predicts the growth rate in transistor density? Are there sufficient incentives for code efficiency when there seem to be surplus compute resources just idly depreciating.
MySQL primary/secondary replication was considered a viable distributed database system, but securing replication depends upon cert exchange and (optionally), PKI, DNS, and IP tunnels of some sort. And then who has root, write to the journal and tables and indexes on the filesystem, UPDATE, and DELETE access in an inter-organizational distributed systems architecture with XML, Web Services, our very own ESB to scale separately from the database replication and off-site backups that nobody ever checks against the online data, and fragmented and varyingly-implemented industry standards that hopefully specify at least a sufficient key for the record that's unique across ledgers/systems/databases.
BitTorrent DHT magnet: links were extant.
Linden Dollars in Linden Lab's Second Life (there's a price floor on land, which is necessary to sell digital assets/goods/products/services)
and accumulated avatar value in e.g. EverQuest and WarCraft (for which there were secondhand markets).
ACH was ACH: GPG-signed files over SFTP on the honor of the audited bank to not allow transfers that deposit money that doesn't exist.
There was no common struct for banking APIs (as apparently only e.g. Plaid, Quicken, and Mint solve for): ledger transactions have a fixed width text field that may contain multiple fields concatenated into one string, and there's no "payee URI" column in the QIF or CSV dumps of an account ledger.
To request more than e.g. the past 90 days of one's own checking account ledger, one was expected to parse tables out of per-month PDFs with e.g. PDFminer at $20 apiece, and then think up ones own natural key in order to merge and lookup records because (2008-01-01,3.99) and (2008-01-01,3.99,storename) are indistinct as a natural key (and when hashed). If you loan a your bank money (for them to now freely invest in the other side, since GLBA in 1999), wouldn't you think that the least they can do is give you `SELECT * WHERE account_id=?` as a free CSV without any datetime limitation in regards to what's offline and what's online.
"Audit the Fed", "Audit DoD" were being chanted by economically-aware citizens amidst severe correction and what was then the most severe recession since the Great Depression: the "Great Recession" it was called, and payouts to essential cronies (who hadn't saved wheat for the famine) were essential.
Overdraft was an error charged to the customer, who didn't build an inconsistent system (CAP theorem) that allows spending money that doesn't exist (at interest charged to the consumer/taxpayer).
"Catch Me If You Can" (2002) described the controls for bank fraud at the time. Why are fees so high?
"Office Space" (1999) described penny-shabing / salami-slicing attack: "fractions of a penny".
"Beverly Hills Ninja" (1997) detailed the story of the Great White Ninja and Tanley! (fistpalm)
"Swordfish" (1999) described a domestic disaster and bank transfers confirmable in seconds.