A lightweight collector/sidecar could be deployed closing to the target monitored system, to collect metrics, traces, and logs.
The fallbacker would do nothing when facing failure data.
yaml