Could entropy harvested from DRAM behavior contribute to decentralized trust scoring?

I’ve been exploring the idea of using DRAM access behavior — specifically memory bandwidth patterns and latency variance — as a way to generate a validator integrity score.
Not for random number generation or consensus replacement, but as a supplemental metric for trust scoring or anomaly detection.

For example:
• Could periodic memory state checks serve as a “heartbeat” to detect hardware spoofing or entropy manipulation?
• Could ZK-SNARKs or MPC attest to hardware-level state ranges without exposing raw memory data?
• Could AI agents (off-chain) flag suspicious behavior by learning “normal” patterns of memory usage per validator?

I’m aware this doesn’t replace coin-flip or commitment schemes, and entropy alone isn’t enough — but could this augment existing cryptographic trust layers?

Would love to hear from anyone who’s worked on similar ideas, especially in:
• zk-based side-channel attestation
• multiparty hardware verification
• entropy-hardening at runtime
• or DRAM-based randomness models

Happy to be proven wrong — or pointed to any research we might be missing.

Edit: Added additional technical details and references in the comments below.