"rndc" system test fails due to rate-limit of built-in "_bind" view
Summary
A series of tests against the built-in "_bind" view sometimes fails due to the rate-limit.
BIND version used
9.18.1
Steps to reproduce
Run the "rndc" system test on a sufficiently fast machine. It is unclear to me what constitutes "sufficiently fast". I've had it succeed on low-end VMs and fail on real hardware.
What is the current bug behavior?
Too many queries against the "_bind" built-in view occur within a second, causing the rate limiting to kick in and some of the tests fail.
What is the expected correct behavior?
The "rndc" system test always succeeds.
Relevant logs and/or screenshots
The log file from the "ns4" nameserver for the "rndc" test. See lines 3353-3356. rndc.ns4.named.run.fail
Possible fixes
I first tried to override the built-in "_bind" view to remove the rate-limit, but this just caused more tests to fail due to the extra unexpected zones.
Eventually I simply introduced artificial delay after the third query. I don't know if this is the correct fix, but it seems to work consistently. rndc.git.patch