"Time to wait for a response from endpoint" will be effective as soon as Tibco Mashery Local (ML) creates a connection to the Backend.
ML will return an error if the Backend doesn't respond within the specified (Time to wait for a response from endpoint) time.

"Seconds to attempt initial connection to endpoint" is the time Mashery Local (ML) will try to create a successful connection to the Backend.

To investigate further, please also refer to Article#000042743 where different time fields captured in the access log are explained.

All Supported Platforms

> ML's timeout counter seems to be starting after it creates a connection to the backend/sends 1st byte
> Client Transfer time is not included.

ML will return "HTTP504 Gateway Timeout" when
-- remote_total_time exceeds the "Time to wait for a response from endpoint"

client_transfer_time:
- is the time taken for transfer of request/response between Client and ML

remote_total_time
- is the time taken for the exchange between ML and Backend;
-- includes time taken for transfer as well Backed processing time

exec_time
- includes remote_total_time plus any processing time ML takes.
 

This article provides inputs for setting the timeout fields for endpoints and also explains how to interpret access log