Regarding privacy... I was downmodded on another thread for stating the obvious, but it is the Pilot's union that does not want longer cockpit voice recordings. The logic is reasonable enough; two hours before an accident should be sufficient to give adequate information for crash investigators. The issue here is that it isn't an accident, it really should be a criminal investigation into the activity in the cockpit.
An airline pilot is a professional, and they don't want to work in an environment where every conversation can be analyzed later, independent of the outcomes.
Efficacy... "it's just metadata." The same reasons we dislike the NSA dragnet is the reason why it is a bad idea for every detail to be recorded and stored indefinitely.
Air transportation is traditionally extremely safe. A very substantial amount of money is put into it to get this outcome. The issue with trying to make marginal improvements is that the return on investment is extremely low.
And back to cost, at $2/message, a message broadcast every 60 seconds on a 6-hour flight with 300 passengers is a premium of $24 per passenger. That would roughly cover position, heading, altitude, and any alarms only. If you wanted to add voice data you are likely looking at something in the range of $60 more per passenger for the flight.
What was needed here was a detachable ELT that activates on impact/submersion and floats on the surface. The logistics of making this sufficiently robust are non trivial, but it would be substantially cheaper than 10 flights with real-time voice streaming from the cockpit, and provide substantially more useful information.