Evaluate switching to using a watch instead of indivdual calls

[eytan-avisror]

Currently we make calls according to the polling configuration:

  configuration:
    successThreshold: 10
    failureThreshold: 10 
    interval: 1s

In this case we would need 10 successful calls to succeed, or 10 failed calls to fail, with 1 second interval.

Instead we could make the configuration more suitable for using watches:

  configuration:
    successDuration: 10s
    failureDuration: 10s

We could make a single call per resource to open a watch, and use time tickers to count the time we are in success/fail state instead.