topics.txt

	- What can go wrong
Use runscope as a proxy whilst you’re developing, and do it in staging to see whats up. Honestly we’ve used it in production for a while too, and had clients retry on the direct URL if runscope was unavailable. This was never an issue but paranoia.
	- Performance
Make the most efficient HTTP calls you can. You might be making HTTP calls with a network library built into the language, or one built by somebody else. Either way, ensure it’s using HTTP/1.1 with keep alive on it, and if possible use a HTTP/2 client library.
HTTP/1.1 keep alive can knock 20% off of repeated connections to the same host depending on how that connection is maintained.
HTTP/2 is baller and async is cool. If they dont use HTTP/2 you can give yourself some with a proxy.
APIs might vanish without a moments notice, maybe ask your backend friends to make a proxy or abstraction layer. Or ensure you have an SLA if its important. Google likes to kill its free APIs and startups can go bust at a moments notice.

Protobuf required is forever or used to be, now yay optional everything
https://developers.google.com/protocol-buffers/docs/pythontutorial

A lightweight debugging/tracing proxy like https://httptoolkit.tech/[HTTP Toolkit] can give you some more insight into what is going on between your local client, which is handy for debugging