Why would you need that many? Trains can keep pretty reasonable track of the distance between two balises using odometers.

There are also other solutions, e.g. LZB uses a phase change of the signal every 100 meters as an additional precision signal.

Because balises cost money, are payed for by infrastructure companies. Their position needs to be decided a priori, so you can't have the whole resolution on the whole network in order to get as close to the physical limits between trains as possible. Also it would be nice to be able to implement automatic shunting without additional infrastructure costs.

There are now hybrid systems for ETCS L2 that allow virtual blocks as short as 1 meter long. Balis + odometer is quite robust (and indeed how most CBTC systems with moving block tech function).

I guess I'm confused by why non track -based solutions are necessary? ETCS L2 would gracefully handle a system restart and automatically force a 40kph speed limit until a train traversed the unknown track proving no occupancy, and allowing proceeding trains to go faster. This all works on a track-based system.

Because railway infrastucture companies would like to not pay for them, not build more tracks and still get more capacity due to "digitization".

Not sure what you're arguing, the line affected by this tragedy was migrating to ETCS L2. Its installation costs are getting lower too, I don't think there's an actual problem here.

I wasn't talking about a specific line at all. Research doesn't need a clear problem to try to find better solutions. Also Deutsche Bahn does think there is enough of a "problem", because they do support that research.