One of the challenges faced by borrow-lend protocols is that depositor a borrower granularity is low. But rate sensitivity tends to be quite high at the margins.

This creates conditions where a single depositor like His Excellency or large borrower like Abraxas can unwind, moving rates substantially in the short term.
While rates often normalize quickly as new “hot money” figures out ways to capitalize on some new equilibrium of rates, it does mean there is a bullwhip effect where small (relative to the entire system) changes in deposit supply or borrow demand create significant volatility in rates.
This has largely just been a normal feature of DeFi markets, where lending and borrowing moves quickly. But with the growing dominance of carry trades in lending market activity, a bullwhip effect in rates can, even if temporary, have major PnL impacts on carry trades with multiple turns of leverage.
I think it is a mistake for DeFi protocols to focus so heavily upon institutional deposit bases, because you end up with a lot of concentration where if that depositor merely tweaks their position at the margins, it can set off a yo-yo kind of path in rates where borrowers bleed cash.
There are several answers to this. The harder one is to actively court a granular, diversified deposit base. This usually means a significant investment in retail marketing and UX, neither of which crypto is very good at, if we’re honest.
Another is rates that don’t update in real time. The original MakerDAO CDP vaults would be manually updated, and typically could go weeks or months without a change. This reduced uncertainty for borrowers who simultaneously knew the rate was semi-fixed in the short run, and could also predict the new rate a week in advance by monitoring the forum.
Painting with a broad brush, I think DeFi lending had stagnated the last few years, with a focus on experimentation in underwriting (beginning with Ethena, Pendle tokens and evolving into looser and looser standards like with Stream, Resolv, Elixir, etc).
Notwithstanding minor experiments in loan structure like 40 Acres (on Base) or 3Jane or Credit (on World Chain), there has been little development beyond simple overnight margin lending lately, which keeps the depositor and borrower universe limited due to a tiny suite of available products,