Reefbane is a unique fishing rod that has a special affix: Glows while in an Area containing a Unique Fish. Most of the research around Reefbane is around determing what this glowy fishing rod can tell us.
Our current operating assumption is that Reefbane does not have any special drop requirements, and is merely mirror-tier rarity. (In fact, if Reefbane is not somehow drop-restricted, it might be the rarest unique item in the game.)
Per the Inscribed Ultimatum Unique Item Tier List, Reefbane is a T0 Unique -- the rarest category.
Sources of drops: We've seen examples of natural Reefbane drops from regular white chests, Lab chests, rare mobs, unique mobs, and more. (There are also deterministic ways to obtain Reefbane, e.g. via Reliquary Keys, cards, using an Ancient Orb on SOTS, buying from an old Forsaken Master for 1 mir, etc. -- but we're interested here in natural drops.)
Areas / locations: We have hard proof of natural Reefbane drops occurring in the following Areas: {Waste Pool, Shipyard, Bazaar, Lab, Bog}. The following locations are also Trustworthy / High-Likelihood, but we do not have hard proof for them: {Tunnel (aka Crystal Ore), Bazaar, Arid Lake, Promenade, Arena}. There are many other casually reported drop locations as well. In short, specific areas (or area tags) do not seem to matter.
A Song of the Sirens can be Ancient Orb'd into a Reefbane in any Area, which implies that Reefbane does not have a special League- or Area-based drop restriction. But one cannot use an Ancient Orb to transform a Reefbane into a Song of the Sirens. This is a one-way recipe that can destroy SOTSes while generating Reefbanes.
We know for certain that Fairgraves' Tricorne is not required for Reefbane to drop. This is a pretty well-known exception in the broader community, too.
We've only really seen multiple Reefbanes drop consistently for characters running minmax'd legacy magic find on Standard with Harvest-horizon'd anointed maps using perfect maximum juicing. This is another testament to its extreme scarcity.
After long investigation, Reefbane's Glow mechanics have been figured out. Reefbane will glow if and only if Area has the area_with_water tag and the (area's seed)mod100 = 0 (in other words, the seed ends with 00).
What where the steps to figure area_with_water tag requirement out?
If you spam enough new instances of a zone, Reefbane eventually glows. Sometimes this requires hundreds of attempts; the probability of occurrence is low (rough estiamtion of ~1%).
However, few zones were found were Reefbane never glowed:
The sample sizes n listed above were the number instance creation attempts by our experimenters. All of these values are pretty substantial. With the assumed probability of 1% for a Reefbane proc, failing to see a proc after 600 trials in a zone that can proc has probability ~0.24% - exceedingly unlikely.
The next obvious research question was: Why does Reefbane seemingly fail to glow here? What is special about these zones versus others?
Here are a few statements that was noticed during the testing:
We selected these zones for testing (Quarry, Mines, etc.) because of a clue left in the Reefbane Reveal video.
ItsYoji's Reefbane Reveal video used substantial GGG footage. One scene compares Reefbane's "Glowing" and "Not Glowing" states.
We realized that this scene may be a hint instead of just a demonstration: the area selected for the "Not Glowing" state may have been an intentional clue. Thus, we began testing zones with tilesets similar to the one shown.
In retrospect the hint seems pretty obvious! There is an all-caps "NO GLOW" message, after all.
For other examples of how crucial hints are in GGG media, check out the discussion on Reddit.
What where the steps to figure the seed requirement out?
During 3.12 Heist league, PoorFishWife noticed that Client.txt started to display the generation seeds of each zone entered. Soon after, it was noticed that few zones, where Reefbane glowed, ended in 00. After that, all that was needed was to try and disprove this claim by observing a counter-example. After the initial n = 623 zones tested and zero counter-examples found, the confidence level of the hypothesis were already higher than p>99%. Since then, zero counter-examples have been found.
This question was asked in the Fishcord:
Is the glow from Reefbane caused by creating an instance with Reefbane equipped/held, or does Reefbane detect an inherent property of certain instances?
The answer is the latter: Reefbane detects a hidden property in certain zone instances, including instances which are created without Reefbane involved. Contrast this statement with instances which inherit character data; for example, when a player creates an endgame map, the game checks character data to determine mods for that endgame map instance (it pulls from the character's sextant data, their atlas status/bonus, etc.).
The reason this is important: a special attribute of certain instances exists independent of Reefbane. Reefbane is a detection tool. Reefbane itself does not cause an instance to be "glowy". The glow procs only in instances which already have a preexisting special attribute.
There were two main ways to test this which occurred to us: (1) party up and have one player create instances, which are then joined by a second player who has Reefbane equipped; (2) drop Reefbane outside of a zone, create a new instance without Reefbane, and then retrieve Reefbane and reenter to see if it procs. We picked the latter method just to have a bigger sample.
We had two players testing independently; following cryptographic tradition, we'll call the players Alice and Bob.
Procedure used:
Trials: n > 250 for Bob; n = {149, 2, 45} for Alice. In all cases Reefbane was observed to glow in instances that were created without Reefbane (or any other items) equipped. These trials took a long time.
Miscellany: Bob tested using a fresh low level character; Alice tested using a high-level character who had completed all storyline + optional quests + maps; Alice tested the first trial batch while wearing other fishing gear. None of these miscellany seemed to have any effect on results.
An additional interesting note: In the last trial, both Bob and Alice each created an instance which proc'd Reefbane's glow within the exact same real-world minute, possibly near-simultaneously. This is very low likelihood. This may imply real-world clocktime matters for Reefbane. (Or it could just be a very unusual coincidence, of course.)