Researching Detection Vectors for SMM Automation

I'm conducting focused research on the specific detection vectors that break large-scale SMM automation (account farms, panel backends). The focus is not on basic fingerprint spoofing, but on sustained session behavior that triggers bans after hours/days of operation.

Looking to connect with a few experienced operators who run at scale (500+ accounts) and have detailed data on:

1. At what point accounts get flagged: upon creation, after first actions, or after sustained activity?
2. Which platform-specific parameters are most volatile (e.g., TikTok's _M fingerprint, Instagram's device_id generation).
3. Whether you've seen patterns related to hardware telemetry drift (like performance.memory trends over time).

In exchange for detailed, technical insights, I can offer a custom analysis of your current fingerprinting setup and generate a batch of alternative profiles tailored to your specific platform and scale for testing purposes.

This is a research exchange. If you're a serious operator with detailed logs and pain points, reply here with your main platform and the scale you operate at. I'll reach out via DM.

Dealing with this right now. Running a small panel, and bots get burned after a few hours—not at creation. Looks like they're tracking something over time.
Tried switching fingerprint services, but they only change the static stuff (canvas, WebGL). The bans kept happening.

OP, if your research finds a way to mimic that 'over-time' behavior (like battery drain or memory use changing), that would be a game changer. Would be willing to test it.

To others here: has anyone found a service that actually solves the sustained activity ban, not just the first check?