First, agencies have too much data. There are solutions to soft through all of that noise but it makes it very difficult to casually catch something you’re not actively searching for.

That's where all the machine learning and data mining research comes in. Want to automatically tell which users are of note? Apply some NLP models for that. Too much footage compared to storage? There are algorithms to determine which sections of the footage are relevant enough to keep. Metadata also helps, but now a lot of research is dedicated from creating some sort of summary metadata for loads of unstructured data.

Second, because everyone is painfully aware of technology able to spy on you now, and most law enforcement relies too heavily on subpoenas of data from these devices rather than active taps, it is far easier to avoid surveillance when you want privacy.

That's just wrong and I have no idea why you think that. In the USA, the government can issue a gag order to prevent the companies from even acknowledging they've been subpoena'd, there've been multiple cases of Amazon or Meta providing data to the police without being coerced into it, and the vast majority of the internet is built on tracking and untrustworthy proprietary software. You might want to look into fingerprinting. It also doesn't matter too much if the data ever gets to the government, as if these corporations couldn't do fishy stuff with your data themselves in case the NSA/FBI/CIA were too much of a hassle.

Every proprietary OS has backdoors implanted into it, Google doesn't even pretend they don't read your emails, and apparently first world countries have company-owned cameras all over the place. Sure, there might be ways to degoogle your phone, use Tor and VPNs in a FOSS-only Linux distribution while self-hosting everything, but that's not necessarily easy by itself, much less in case you also still want to remain connected to normal people who use WhatsApp.