[danieljbruce]

This PR is not meant to be merged, just posted for debugging. A diff in this PR is shown against the branch that has the mock server set up for testing retries. This diff shows the new test which tests retries against the Gapic layer, ignoring the handwritten layer entirely which will be useful for isolating problems. Also, it sets server streaming retries to true in the gapic layer.

One possible strategy for isolating problems is to modify the const options: CallOptions = { line passing retry options directly into a call to the Gapic layer.

[leahecole]

Suggested change

	const backOffSettings = {
	initialRetryDelayMillis: 10,
	retryDelayMultiplier: 2,
	maxRetryDelayMillis: 60000,
	};
	const shouldRetryFn = function checkRetry(error: GoogleError) {
	return false; // return [14, 4].includes(error!.code!);
	};
	const retry = new RetryOptions([], backOffSettings, shouldRetryFn);
	const options: CallOptions = {
	otherArgs: {
	header: {
	'bigtable-attempt': 0,
	},
	},
	retry,
	maxRetries: 3,
	const backOffSettings = {
	initialRetryDelayMillis: 10,
	retryDelayMultiplier: 2,
	maxRetryDelayMillis: 60000,
	maxRetries: 3
	};
	const shouldRetryFn = function checkRetry(error: GoogleError) {
	return false; // return [14, 4].includes(error!.code!);
	};
	const retry = new RetryOptions([], backOffSettings, shouldRetryFn);
	const options: CallOptions = {
	otherArgs: {
	header: {
	'bigtable-attempt': 0,
	},
	},
	retry,

[leahecole]

this looks uglier than it is - it just moves maxRetries to the backoffSettings