After making some optimizations, We discovered a weird problem with prev/next buttons on AB’s site. I realized what the problem was and remembered I had already put a bandaid on it before but forgot about it when I made the optimizations. I wrote the following note above the function I had to fix.

/**
 * populateIDsIfDNE
 *
 * Hey future smart person
 * Before you decide to optimize by lowering PREFERENCE_WINGSPAN_PREV_NEXT, thinking it will reduce the time
 * when populating prev/next ids in TABLE `search_queries_backend_ids`, just remember we have to
 * account for maintaining list offsets while editing the very list we are on.
 * Similar to the comment in commit 695aacfd96df0dbc1ff82af924c8e399249d656a
 * We have an offset problem for any given search if we get to the last id in TABLE `search_queries_backend_ids`
 *
 * Here are the steps that cause a problem:
 *
 * * load list of units to check https://ab/en/system/units/?outofdate=1
 * * view a unit in the list (at which time the prev/next units are saved in TABLE `search_queries_backend_ids`)
 * * save the unit as checked (whereby it will not show up in original search)
 * * (optional) click next and edit next unit
 * * (optional) click next and do not edit next unit
 * * LET の be the number of units edited during this process
 * * click next until we get to last offset (at end of PREFERENCE_WINGSPAN_PREV_NEXT)
 * * this triggers a reload of unit ids (during which any checked units will not be found because they no longer need to be checked)
 * * let ォ be $offset_of_current at that time
 * * Now in TABLE `search_queries_backend_ids`, the new first unit was the のth unit in the original list
 * * click [next unit] on unit edit page (goes to offset = PREFERENCE_WINGSPAN_PREV_NEXT in TABLE `search_queries_backend_ids`)
 * * discover units were skipped.  How many? Let's see.. オ+の!
 *
 * The オ の variable values might not be defined or used exactly as outlined above,
 * but hopefully this explains the problem with short wingspans
 *
 * Good luck,
 * Rob
 * March 4th, 2020
 *
 */

After hotfixing the above problem, I got an email while running tests on AB

PHP Notice:  Constant SITE_VERSION already defined in /var/www/ab/versions/2020-03-03-2223/prepend.php on line 3

Wait, what?

I had to issue the following world’s funniest hotfix version hotfix

diff --git a/prepend.php b/prepend.php
index 530e1ab9..13f25731 100644
--- a/prepend.php
+++ b/prepend.php
@@ -1,6 +1,5 @@
 <?php
-define('SITE_VERSION','4.5.9');
-define('SITE_VERSION','4.5.9a');
+define('SITE_VERSION','4.5.9b');

That is all.