The remaining population of COVID patients is a good control group. Or, if you prefer, randomly select sub-populations with similar demographic statistics.
Yes, let's throw out valuable field evidence just because it doesn't fit some narrow definition of "valid experiment".
Compare 700 people who received a treatment to 700 demographically similar people who received the standard treatment in the same or similar hospital and you're going to get a useful result.
We can do your fancy RCT later when we aren't dying en masse from a novel and highly lethal virus.
I mean, people have tried doing studies like that for CQ/HCQ. They are riddled with questionable assumptions (that patient groups are similar or treatment regimens are equivalent) that make their conclusions completely unreliable.
In the end we will most likely find that all positive anecdotal evidence found was just confirmation bias. That's why RCTs are so important.
