Consider the interaction between tx_confirmation and clear_interval parameters

[seanchen1991]

In #1951, the interaction between the tx_confirmation and clear_interval parameters came up quite a bit.

In summary, one of the auxiliary effects of the tx_confirmation parameter is to control whether pending tx's are re-submitted when they are not successfully delivered. An important caveat with this capability is that the clear_interval parameter must be set to 0 for this to occur:

# in this case, pending tx's are queued up for re-submission
clear_interval = 0
tx_confirmation = true

In the situation when clear_interval is greater than 0 such that pending packets are eagerly cleared once the interval (which is counted based on the number of blocks) has been reached, the re-submission of pending packets under the purview of the tx_confirmation parameter is overridden.

In other words, pending packets are cleared regardless when clear_interval > 0, and they are cleared when clear_interval = 0 and tx_confirmation = true. Hence, the only situation when they are not being cleared is when clear_interval = 0 and tx_confirmation = false.

Given all of this context, the question is: should this interaction be more straightforward? While #1951 attempted to document the interaction between these two parameters (frankly not in a particularly obvious way), it's a pretty nuanced point that will likely be glossed over by operators. The best way to address this is to come up with a more straightforward scheme that doesn't rely on the interplay between two distinct parameters.

This issue merely seeks to document this interaction and ask whether there is a better way to structure this behavior.

[adizere]

One option is to completely remove the tx_confirmation option. Instead of that option, we would infer if we should do confirmation iff telemetry is enabled. So telemetry enabled = true implies tx_confirmation = true.