Saw failure for first time ever -- one of the transport_test-scripted tight timeouts was a bit too tight for the GitHub runners again. Increased by a few microseconds again.#80