diff --git a/apps/zbugs/docker/init_upstream/init.sql b/apps/zbugs/docker/init_upstream/init.sql
index 478dad2dc..e292e0c5a 100644
--- a/apps/zbugs/docker/init_upstream/init.sql
+++ b/apps/zbugs/docker/init_upstream/init.sql
@@ -7,6 +7,8 @@ DROP TABLE IF EXISTS "user",
 "userPref",
 "zero.schemaVersions" CASCADE;
 
+-- user
+
 CREATE TABLE "user" (
     "id" VARCHAR PRIMARY KEY,
     "login" VARCHAR NOT NULL,
@@ -19,19 +21,51 @@ CREATE TABLE "user" (
 CREATE UNIQUE INDEX user_login_idx ON "user" (login);
 CREATE UNIQUE INDEX user_githubid_idx ON "user" ("githubID");
 
+-- issue
+
 CREATE TABLE issue (
     "id" VARCHAR PRIMARY KEY,
     "shortID" INTEGER GENERATED BY DEFAULT AS IDENTITY (START WITH 3000),
-    "title" VARCHAR NOT NULL,
+    "title" VARCHAR(128) NOT NULL,
     "open" BOOLEAN NOT NULL,
     "modified" double precision DEFAULT (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000),
     "created" double precision DEFAULT (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000),
     "creatorID" VARCHAR REFERENCES "user"(id) NOT NULL,
     "assigneeID" VARCHAR REFERENCES "user"(id),
-    "description" TEXT DEFAULT '',
+    -- Size chosen because max we currently have in legacy data is ~9KB.
+    "description" VARCHAR(10240) DEFAULT '',
     "visibility" VARCHAR DEFAULT 'public' NOT NULL
 );
 
+
+CREATE OR REPLACE FUNCTION update_modified_column()
+RETURNS TRIGGER AS $$
+BEGIN
+    NEW.modified = (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000);
+    RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
+
+CREATE TRIGGER issue_set_last_modified
+BEFORE INSERT OR UPDATE ON issue
+FOR EACH ROW
+EXECUTE FUNCTION update_modified_column();
+
+CREATE OR REPLACE FUNCTION issue_set_created_on_insert()
+RETURNS TRIGGER AS $$
+BEGIN
+    NEW.created = (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000);
+    RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
+
+CREATE TRIGGER issue_set_created_on_insert_trigger
+BEFORE INSERT ON issue
+FOR EACH ROW
+EXECUTE FUNCTION issue_set_created_on_insert();
+
+-- viewState
+
 CREATE TABLE "viewState" (
     "userID" VARCHAR REFERENCES "user"(id) ON DELETE CASCADE,
     "issueID" VARCHAR REFERENCES issue(id) ON DELETE CASCADE,
@@ -39,6 +73,8 @@ CREATE TABLE "viewState" (
     PRIMARY KEY ("userID", "issueID")
 );
 
+-- comment
+
 CREATE TABLE comment (
     id VARCHAR PRIMARY KEY,
     "issueID" VARCHAR REFERENCES issue(id) ON DELETE CASCADE,
@@ -75,17 +111,46 @@ BEFORE INSERT ON comment
 FOR EACH ROW
 EXECUTE FUNCTION comment_set_created_on_insert();
 
+CREATE OR REPLACE FUNCTION validate_comment_body_length()
+RETURNS TRIGGER AS $$
+BEGIN
+  IF NEW.body IS NOT NULL THEN
+    -- The launch post has a special case maxlength of 1024 because trolls
+    IF NEW."issueID" = 'duuW9Nyj5cTNLlimp9Qje' AND LENGTH(NEW.body) > 1024 THEN
+      RAISE EXCEPTION 'Column value exceeds maximum allowed length of %', 1024;
+    END IF;
+    -- Length chosen because we have some old comments that are ~44KB.
+    IF LENGTH(NEW.body) > 64*1024 THEN
+      RAISE EXCEPTION 'Column value exceeds maximum allowed length of %', 64*1024;
+    END IF;
+  END IF;
+  RETURN NEW;
+END;
+$$ LANGUAGE plpgsql;
+
+CREATE TRIGGER check_comment_body_length
+BEFORE INSERT OR UPDATE ON comment
+FOR EACH ROW
+EXECUTE FUNCTION validate_comment_body_length();
+
+
+-- label
+
 CREATE TABLE label (
     "id" VARCHAR PRIMARY KEY,
     "name" VARCHAR NOT NULL
 );
 
+-- issueLabel
+
 CREATE TABLE "issueLabel" (
     "labelID" VARCHAR REFERENCES label(id),
     "issueID" VARCHAR REFERENCES issue(id) ON DELETE CASCADE,
     PRIMARY KEY ("labelID", "issueID")
 );
 
+-- emoji
+
 CREATE TABLE emoji (
     "id" VARCHAR PRIMARY KEY,
     "value" VARCHAR NOT NULL,
@@ -181,6 +246,8 @@ BEGIN
 END;   
 $$ LANGUAGE plpgsql;
 
+-- userPref
+
 CREATE TABLE "userPref" (
     "key" VARCHAR NOT NULL,
     "value" VARCHAR NOT NULL,
@@ -189,6 +256,8 @@ CREATE TABLE "userPref" (
     PRIMARY KEY ("userID", "key")
 );
 
+-- zero.schemaVersions
+
 CREATE SCHEMA IF NOT EXISTS zero;
 
 CREATE TABLE IF NOT EXISTS zero."schemaVersions" (
@@ -206,34 +275,6 @@ INSERT INTO zero."schemaVersions" ("lock", "minSupportedVersion", "maxSupportedV
 VALUES (true, 3, 5) ON CONFLICT DO NOTHING;
 
 
--- last modified function and trigger
-CREATE OR REPLACE FUNCTION update_modified_column()
-RETURNS TRIGGER AS $$
-BEGIN
-    NEW.modified = (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000);
-    RETURN NEW;
-END;
-$$ LANGUAGE plpgsql;
-
-CREATE TRIGGER issue_set_last_modified
-BEFORE INSERT OR UPDATE ON issue
-FOR EACH ROW
-EXECUTE FUNCTION update_modified_column();
-
-CREATE OR REPLACE FUNCTION issue_set_created_on_insert()
-RETURNS TRIGGER AS $$
-BEGIN
-    NEW.created = (EXTRACT(EPOCH FROM CURRENT_TIMESTAMP) * 1000);
-    RETURN NEW;
-END;
-$$ LANGUAGE plpgsql;
-
-CREATE TRIGGER issue_set_created_on_insert_trigger
-BEFORE INSERT ON issue
-FOR EACH ROW
-EXECUTE FUNCTION issue_set_created_on_insert();
-
-
 COPY "user"
 FROM
     '/docker-entrypoint-initdb.d/users.csv' WITH CSV HEADER;
@@ -254,6 +295,7 @@ COPY "comment"
 FROM
     '/docker-entrypoint-initdb.d/comments.csv' WITH CSV HEADER;
 
+
 -- Create the indices on upstream so we can copy to downstream on replication.
 -- We have discussed that, in the future, the indices of the Zero replica
 -- can / should diverge from the indices of the upstream. This is because
diff --git a/apps/zbugs/docker/init_upstream/issues.csv b/apps/zbugs/docker/init_upstream/issues.csv
index ed7063b06..c5cb6da97 100644
--- a/apps/zbugs/docker/init_upstream/issues.csv
+++ b/apps/zbugs/docker/init_upstream/issues.csv
@@ -3055,7 +3055,7 @@ Profile attached. More information when less tired :).
 
 [Profile-live.json.zip](https://github.com/rocicorp/reflect-server/files/9663620/Profile-live.json.zip)
 "
-N983dAQkVWxJwlWi8nb_N,75,We can get rid of the whole complicated updateClients (which was needed due to async hashing and indexeddb).  Callers can use setClients (and they need to put their own chunks).,False,1677091437000.0,1664205709000.0,nqYkxAGMnzk7Y5STjZryV,,"        We can get rid of the whole complicated updateClients (which was needed due to async hashing and indexeddb).  Callers can use setClients (and they need to put their own chunks).  
+N983dAQkVWxJwlWi8nb_N,75,We can get rid of the whole complicated updateClients (which was needed due to async hashing and indexeddb)..,False,1677091437000.0,1664205709000.0,nqYkxAGMnzk7Y5STjZryV,,"Callers can use setClients (and they need to put their own chunks).
 
 Fine to do as a follow up as this is already quite a big pr.
 
@@ -3365,10 +3365,10 @@ zrpBifgcCO7g1DjYQ1v9A,90,"Rails ""done done""",False,1677091445000.0,16570873140
 - [ ] Update ReplidrawDO to use
 - [ ] Update Repliear to use
 - [ ] Rename to something else... replicache-crud?"
-jC2ZSB7TCGAhDeK_CTNuj,147,Maybe call this test `crossClientGetMutationID` or something to make more clear why it only runs under DD31. A comment would also help.,False,1677091452000.0,1656407388000.0,nqYkxAGMnzk7Y5STjZryV,,"Maybe call this test `crossClientGetMutationID` or something to make more clear why it only runs under DD31. A comment would also help.
+jC2ZSB7TCGAhDeK_CTNuj,147,Maybe call this test `crossClientGetMutationID` or something to make more clear why it only runs under DD31.,False,1677091452000.0,1656407388000.0,nqYkxAGMnzk7Y5STjZryV,,"A comment would also help.
 
 _Originally posted by @aboodman in https://github.com/rocicorp/replicache-internal/pull/177#discussion_r906419411_"
-bbBCilPYtx_FfumPLQRqG,148,"Why not adjust these tests to have a dag.Read. Even if most of the commit types / tests don't need it, some do. Would be easier to read the test if you didn't have to understand why it's OK to pass null here.",False,1677091453000.0,1656407353000.0,nqYkxAGMnzk7Y5STjZryV,,"Why not adjust these tests to have a dag.Read. Even if most of the commit types / tests don't need it, some do. Would be easier to read the test if you didn't have to understand why it's OK to pass null here.
+bbBCilPYtx_FfumPLQRqG,148,"Why not adjust these tests to have a dag.Read. Even if most of the commit types / tests don't need it, some do.",False,1677091453000.0,1656407353000.0,nqYkxAGMnzk7Y5STjZryV,,"Would be easier to read the test if you didn't have to understand why it's OK to pass null here.
 
 _Originally posted by @aboodman in https://github.com/rocicorp/replicache-internal/pull/177#discussion_r906418791_"
 0aANjS4hkfgYM6jyk0bnz,111,DD 3.1,False,1709536712000.0,1655804122000.0,nqYkxAGMnzk7Y5STjZryV,,"Tracking bug for https://www.notion.so/replicache/DD-3-1-e42489fc2e6b4340a01c7fa0de353a30