There’s a big difference between the metadata that the Snowden leaks are talking about and how they’re used and the metadata the signal server (or its subcontracted provider) has and how and under what conditions it’s able to be used.

The metadata that is the subject of the statement “we kill people based off metadata” is unencrypted cell phone signals and other broadly plaintext requests sent over a system that by design also includes location telemetry. That information could be easily obtained en masse through a man in the middle attack or through the lawful intercept backdoors built into the equipment that carries the information itself (which is less of a man in the middle attack and more of a man in the middle design).

This is different from the signal metadata both in form and content. The signal metadata is not vulnerable to a mitm attack and the agencies implicated in the Snowden leaks would have to actually go through the legal hoopla required in order to get just the metadata itself. Same as they would have to if they wanted the actual content.

Amazon does comply with law enforcement requests often without requiring a warrant, but the difference between requiring a request be made as opposed to simply being able to collect that metadata freely and package it as actionable intelligence is significant.

All messaging systems are vulnerable to this attack. If you send or receive a message then you, the other party and any intermediary like a server are subject to the laws of the places they’re physically located.

Again, I’m not arguing, I’m trying to make this very convoluted system clearer.