I think that it's a pretty reasonable payout - people running bug bounties are mostly interested in actual security impact and practical exploitability. And the necessity to hit G server thousands of times to extract even a short video reduces that by quite a large proportion; which is definitely not a shot at the researcher, it's a great find, but exploitability is definitely lower than more direct form of IDOR (e.g. input private video ID and get the whole video in response).
By the same logic, blind SQLi will typically be valued 'less' (hence pay out less) than SQLi with output.
By the same logic, blind SQLi will typically be valued 'less' (hence pay out less) than SQLi with output.