diff --git a/src/config.py b/src/config.py
index d7f1d30d..ba5d7d7a 100644
--- a/src/config.py
+++ b/src/config.py
@@ -80,13 +80,6 @@ class Config(BaseSettings):
     TELEGRAM_API_HASH: str | None = None
     TELEGRAM_SESSION_STRING: str | None = None
 
-    # @model_validator(mode="after")
-    # def validate_sentry_non_local(self) -> "Config":
-    #     if self.ENVIRONMENT.is_deployed and not self.SENTRY_DSN:
-    #         raise ValueError("Sentry is not set")
-
-    #     return self
-
 
 settings = Config()
 
diff --git a/src/database.py b/src/database.py
index c15f5787..8979eaae 100644
--- a/src/database.py
+++ b/src/database.py
@@ -697,13 +697,9 @@ def _is_transient_connection_error(exc: BaseException) -> bool:
     return _is_stale_connection_error(exc) or _is_concurrent_use_error(exc)
 
 
-# Shared retry budget for stale / concurrent-use connection errors.
-# Three retries (25 ms + 50 ms + 100 ms backoff) give the event loop enough
-# time to dispose dirty connections under burst webhook concurrency.
-# Under traffic spikes the LIFO pool can hand out several consecutively
-# dirty connections; 2 retries proved insufficient (9 Sentry events on
-# 2026-04-12).  Bumping to 3 retries with wider backoff makes exhaustion
-# astronomically unlikely while adding only ~175 ms worst-case latency.
+# The LIFO pool can return consecutive stale/concurrent-use connections under load.
+# Two retries proved insufficient (9 Sentry events on 2026-04-12); three retries
+# allow 25 + 50 + 100 ms of backoff before propagating the failure.
 _TRANSIENT_MAX_RETRIES = 3
 
 
diff --git a/src/flows/crossposting/meme.py b/src/flows/crossposting/meme.py
index 7e0d81a5..3488f8d9 100644
--- a/src/flows/crossposting/meme.py
+++ b/src/flows/crossposting/meme.py
@@ -136,11 +136,6 @@ def _get_ru_caption_for_crossposting_meme(meme: MemeData, channel: Channel) -> s
     cta = random.choice(CTAS)
     ref_link = "https://t.me/ffmemesbot?start=sc_{}_{}".format(meme.id, channel.value)
 
-    # emoji = get_random_emoji()
-    # referral_html = f"""{emoji} <i><a href="{ref_link}">Fast Food Memes</a></i>"""
-    # caption = escape(meme.caption, quote=False) if meme.caption else ""
-    # text = caption + "\n\n" + referral_html
-
     text = cta + ": " + f"""<a href="{ref_link}">@ffmemesbot</a>"""
 
     return text
@@ -220,8 +215,7 @@ async def post_meme_to_tgchannelen():
     next_meme = MemeData(**meme_data)
     logger.info(f"Next meme for TG Channel EN: {next_meme.id}")
 
-    # Persist the ranker decision for retro analysis. Failure must NOT propagate
-    # — Prefect retry would republish the meme.
+    # Ranker logging is best effort so analytics cannot block posting.
     if decision:
         try:
             await log_ranker_decision(**decision)
@@ -280,8 +274,7 @@ async def post_meme_to_tgchannelru():
     next_meme = MemeData(**meme_data)
     logger.info(f"Next meme for TG Channel RU: {next_meme.id}")
 
-    # Persist the ranker decision for retro analysis. Failure must NOT propagate
-    # — Prefect retry would republish the meme.
+    # Ranker logging is best effort so analytics cannot block posting.
     if decision:
         try:
             await log_ranker_decision(**decision)
diff --git a/src/flows/storage/describe_memes.py b/src/flows/storage/describe_memes.py
index 284e1493..ff1f9c5c 100644
--- a/src/flows/storage/describe_memes.py
+++ b/src/flows/storage/describe_memes.py
@@ -70,7 +70,6 @@ async def describe_single_meme(
     file_id = meme_row["telegram_file_id"]
     existing_ocr = meme_row["ocr_result"] or {}
 
-    # Download image from Telegram
     try:
         image_bytes = await download_meme_content_from_tg(file_id)
     except Exception as e:
@@ -80,7 +79,6 @@ async def describe_single_meme(
 
     image_b64 = base64.b64encode(image_bytes).decode("utf-8")
 
-    # Call vision model
     try:
         result = await call_openrouter_vision(image_b64, log, deadline=deadline, keys=keys)
     except Exception as e:
diff --git a/src/localizer.py b/src/localizer.py
index 28311c93..3c2de5c9 100644
--- a/src/localizer.py
+++ b/src/localizer.py
@@ -3,7 +3,6 @@
 
 import yaml
 
-# not sure where to put this const
 DEFAULT_LANG = "en"
 
 ALMOST_CIS_LANGUAGES = [
diff --git a/src/main.py b/src/main.py
index 0ddc9f12..485c53f9 100644
--- a/src/main.py
+++ b/src/main.py
@@ -19,21 +19,16 @@
 
 @asynccontextmanager
 async def lifespan(_application: FastAPI) -> AsyncGenerator:
-    # Startup
     tgbot_app.application = tgbot_app.setup_application(settings.ENVIRONMENT.is_deployed)
     await tgbot_app.application.initialize()
 
-    # flush all redis keys on startup for debug
-    # await redis.redis_client.flushall()
-
-    # if is_webhook:  # all gunicorn workers will call this and hit rate limit
-    #     await bot.setup_webhook(bot.application)
+    # Do not register the webhook here: every Gunicorn worker runs lifespan,
+    # so startup registration would repeatedly hit Telegram's rate limit.
 
     yield
 
     if settings.ENVIRONMENT.is_testing:
         return
-    # Shutdown
     await redis.pool.disconnect()
 
 
diff --git a/src/storage/etl.py b/src/storage/etl.py
index 5b4ed540..1e89bf25 100644
--- a/src/storage/etl.py
+++ b/src/storage/etl.py
@@ -55,10 +55,6 @@ async def insert_parsed_posts_from_telegram(
     *,
     discover_candidates: bool = True,
 ) -> None:
-    # 1. find which memes are already in the database
-    # 2. update existing memes
-    # 3. insert new memes
-
     result = await fetch_all(
         select(meme_raw_telegram.c.post_id)
         .where(meme_raw_telegram.c.meme_source_id == meme_source_id)
@@ -202,10 +198,6 @@ async def etl_memes_from_raw_telegram_posts(
     *,
     fresh_only: bool = True,
 ) -> None:
-    # get transformed posts
-    # find ones that are already in the database
-    # create rows and update rows
-    #
     # Engagement filters before raw posts enter the meme table:
     # 1. From each source's latest 10 posts, only promote the top 5 by views.
     #    Raw rows are still stored, so this threshold can be replayed later.
@@ -325,15 +317,6 @@ async def etl_memes_from_raw_telegram_posts(
         },
     )
 
-    # find rows which already exist in db by two index columns:
-    # meme_source_id and raw_meme_id
-    # so we can update existing rows and create new ones
-
-    # join two tables meme_raw_telegram and meme to get
-    # the meme_source_id and raw_meme_id pairs
-    # which are not present in memes table.
-    # That will indicate that we need to create new rows
-
     memes_not_in_memes_table = await fetch_all(
         text(
             """
diff --git a/src/storage/upload.py b/src/storage/upload.py
index 0e071d3b..97839998 100644
--- a/src/storage/upload.py
+++ b/src/storage/upload.py
@@ -78,7 +78,7 @@ async def _image_hash(meme_id: int, meme_type: MemeType, content: bytes, media:
 async def _upload_meme_content_to_tg(
     meme_id: int,
     meme_type: MemeType,
-    content: bytes,  # ??
+    content: bytes,
 ) -> dict[str, Any] | None:
     file_id = None
     media = None
@@ -166,8 +166,6 @@ async def upload_meme_content_to_tg(
     *,
     observability_context: dict[str, Any] | None = None,
 ) -> dict[str, Any] | None:
-    # just a wrapper to handle retries
-
     meme_result = None
     max_attempts = 3
     content_size = len(content)
diff --git a/src/storage/watermark.py b/src/storage/watermark.py
index db4c35e7..2adc7ee4 100644
--- a/src/storage/watermark.py
+++ b/src/storage/watermark.py
@@ -16,10 +16,8 @@ def draw_text_with_outline(
     outline_colour: tuple,
 ) -> None:
     x, y = position
-    # Draw outline
     for adj in range(-1, 2):
         for ops in range(-1, 2):
-            # Avoid the center pixel
             if adj != 0 or ops != 0:
                 draw.text((x + adj, y + ops), text, font=font, fill=outline_colour)
     draw.text(position, text, font=font, fill=text_colour)
@@ -42,39 +40,27 @@ def find_least_detailed_corner(img: Image, text_bbox: tuple, margin: int) -> tup
     - A tuple (x, y) representing the top-left coordinate
         of the selected corner for text placement.
     """
-    # Convert the original image to grayscale for intensity analysis
     gray_img = img.convert("L")
     img_w, img_h = img.size
-    # Calculate the width and height of the text bounding box
     text_width = text_bbox[2] - text_bbox[0]
     text_height = text_bbox[3] - text_bbox[1]
-    # Define dictionary with the coordinates of each corner with some margin applied
     corners = {
         "top_left": (margin, margin),
         "top_right": (img_w - text_width - margin, margin),
         "bottom_left": (margin, img_h - text_height - margin),
         "bottom_right": (img_w - text_width - margin, img_h - text_height - margin),
     }
-    # Initialize with infinity for comparison
     least_detailed_score = float("inf")
-    # To store the x, y coordinates for watermark
     selected_corner = None
     selected_wm_colour = None
-    # Iterate through each corner to find the one with the least detail background
     for _, (x, y) in corners.items():
-        # Crop the image around the current corner
-        # according to the text bounding box size
         sector = gray_img.crop((x, y, x + text_width, y + text_height))
-        # Calculate mean intensity of the cropped area
         mean_intensity = ImageStat.Stat(sector).mean[0]
-        # Determine how close the mean intensity is to either black (0) or white (255)
-        # indicates how uniform the background of the selected sector is
+        # Use mean brightness near black or white as a cheap proxy for a quiet background.
         distance_to_extreme = min(abs(0 - mean_intensity), abs(255 - mean_intensity))
-        # Update if this corner has a lower score (closer to black or white)
         if distance_to_extreme < least_detailed_score:
             least_detailed_score = distance_to_extreme
             selected_corner = (x, y)
-            # Selecting watermark colour based on mean intensity
             selected_wm_colour = (0, 0, 0) if mean_intensity > 178 else (255, 255, 255)
 
     return selected_corner, selected_wm_colour
@@ -110,7 +96,6 @@ def draw_corner_watermark(
         text_position, text_colour = find_least_detailed_corner(base, text_bbox, margin)
         outline_colour = (0, 0, 0, 255) if text_colour == (255, 255, 255) else (255, 255, 255, 255)
         draw_text_with_outline(d, text_position, text, font, text_colour, outline_colour)
-        # text opacity
         txt.putalpha(txt.getchannel("A").point(lambda x: x * text_opacity))
         return Image.alpha_composite(base, txt).convert("RGB")
 
diff --git a/src/tgbot/handlers/admin/forward_channel.py b/src/tgbot/handlers/admin/forward_channel.py
index a91fdd84..ae09c93c 100644
--- a/src/tgbot/handlers/admin/forward_channel.py
+++ b/src/tgbot/handlers/admin/forward_channel.py
@@ -70,10 +70,6 @@ async def handle_forwarded_from_tgchannelru(
 
         await update.message.reply_text(f"Forwarding post to {len(users)}")
         await log(f"Forwarding post to {len(users)}", context.bot)
-
-        # return await update.message.reply_text(
-        #     f"Can't parse caption: {escape(str(update.message.to_dict()))}"
-        # )
     else:
         url = urls[0]
         meme_id_str = url.replace("https://t.me/ffmemesbot?start=sc_", "").split("_")[0]
diff --git a/src/tgbot/handlers/admin/user_info.py b/src/tgbot/handlers/admin/user_info.py
index a057942b..d28b531c 100644
--- a/src/tgbot/handlers/admin/user_info.py
+++ b/src/tgbot/handlers/admin/user_info.py
@@ -71,10 +71,5 @@ async def delete_user_data_confirmation_page(
 
 async def delete_user_data(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
     """Deletes all user data we have for testing purposes"""
-    # user = await get_user_info(update.effective_user.id)
-    # if user["type"] != UserType.ADMIN:
-    #     return
-
-    # TODO: "are you sure" button + callback
     await delete_user(update.effective_user.id)
     await update.effective_user.send_message("Ciao 👋\n\n --> press /start to try again 👾")
diff --git a/src/tgbot/handlers/deep_link.py b/src/tgbot/handlers/deep_link.py
index 4d88a079..a50f0f85 100644
--- a/src/tgbot/handlers/deep_link.py
+++ b/src/tgbot/handlers/deep_link.py
@@ -29,7 +29,6 @@ async def handle_invited_user(
 
     invitor_user_id = share_link.sharer_user_id
 
-    # get invitor user
     invitor_user = await get_user_by_id(invitor_user_id)
     if not invitor_user:
         return  # Invitor doesn't exist
@@ -45,7 +44,6 @@ async def handle_invited_user(
 was already invited by #{invited_user["inviter_id"]}"""
         )
 
-    # set inviter id
     await update_user(invited_user["id"], inviter_id=invitor_user_id)
 
     # Don't reward if the inviter cannot receive the bot notification.
diff --git a/src/tgbot/handlers/upload/upload_meme.py b/src/tgbot/handlers/upload/upload_meme.py
index 2d793c04..c63fd961 100644
--- a/src/tgbot/handlers/upload/upload_meme.py
+++ b/src/tgbot/handlers/upload/upload_meme.py
@@ -271,7 +271,6 @@ async def handle_message_with_meme(update: Update, context: ContextTypes.DEFAULT
         raise Exception(f"not supported format: {update.message.to_json()}")
 
 
-# callback: RULES_ACCEPTED_CALLBACK_DATA
 async def handle_rules_accepted_callback(
     update: Update,
     _: ContextTypes.DEFAULT_TYPE,
@@ -292,7 +291,6 @@ async def handle_rules_accepted_callback(
     )
 
 
-# callback_data = LANG_SETTINGS_END_CALLBACK_DATA
 async def handle_meme_upload_lang_other(
     update: Update,
     _: ContextTypes.DEFAULT_TYPE,
@@ -304,7 +302,6 @@ async def handle_meme_upload_lang_other(
     )
 
 
-# callback: RULES_ACCEPTED_CALLBACK_DATA
 async def handle_meme_upload_lang_selected(
     update: Update,
     context: ContextTypes.DEFAULT_TYPE,
@@ -320,8 +317,6 @@ async def handle_meme_upload_lang_selected(
 
     await log(f"""📥 Meme {meme["id"]} uploaded by #{update.effective_user.id}""", context.bot)
 
-    # TODO: create a meme object from meme_raw_upload
-
     await _edit_upload_callback_caption(
         update.callback_query.message,
         localizer.t("upload.submitted", user_info["interface_lang"]),
diff --git a/src/tgbot/repo/languages.py b/src/tgbot/repo/languages.py
index 2b1b9aad..d03e11ba 100644
--- a/src/tgbot/repo/languages.py
+++ b/src/tgbot/repo/languages.py
@@ -33,8 +33,6 @@ async def add_user_languages(
     user_id: int,
     language_codes: Sequence[str],
 ) -> None:
-    # Prepare a list of dictionaries where each dictionary represents
-    # the values to be inserted for one row.
     values_to_insert = [
         {"user_id": user_id, "language_code": language_code} for language_code in language_codes
     ]
diff --git a/src/tgbot/repo/popups.py b/src/tgbot/repo/popups.py
index 3199ebde..4a3ad6af 100644
--- a/src/tgbot/repo/popups.py
+++ b/src/tgbot/repo/popups.py
@@ -50,11 +50,12 @@ async def update_user_popup_log(
         user_popup_logs.update()
         .where(user_popup_logs.c.user_id == user_id)
         .where(user_popup_logs.c.popup_id == popup_id)
-        .where(user_popup_logs.c.reacted_at.is_(None))  # not sure abot that
+        # Only the first click may claim the reaction.
+        .where(user_popup_logs.c.reacted_at.is_(None))
         .values(reacted_at=datetime.utcnow())
     )
     res = await execute(update_query)
     reaction_is_new = res.rowcount > 0
     if not reaction_is_new:
         logging.warning(f"User {user_id} already reacted to popup {popup_id}!")
-    return reaction_is_new  # so I can filter double clicks
+    return reaction_is_new
diff --git a/src/tgbot/senders/alerts.py b/src/tgbot/senders/alerts.py
index 2c7a3ffc..b3db2d5b 100644
--- a/src/tgbot/senders/alerts.py
+++ b/src/tgbot/senders/alerts.py
@@ -1,8 +1,6 @@
 from telegram import Bot, Message
 
 from src import localizer
-
-# from src.tgbot.senders.keyboards import queue_empty_alert_keyboard
 from src.tgbot.user_info import get_user_info
 
 
@@ -11,5 +9,4 @@ async def send_queue_preparing_alert(bot: Bot, user_id: int) -> Message:
     return await bot.send_message(
         chat_id=user_id,
         text=localizer.t("service.out_of_memes", user_info["interface_lang"]),
-        # reply_markup=queue_empty_alert_keyboard(),
     )
diff --git a/src/tgbot/senders/keyboards.py b/src/tgbot/senders/keyboards.py
index 266b6f49..386cebc5 100644
--- a/src/tgbot/senders/keyboards.py
+++ b/src/tgbot/senders/keyboards.py
@@ -21,8 +21,6 @@
 )
 from src.tgbot.sharing import build_meme_share_button
 
-# IDEA: use sometimes another emoji pair like 🤣/🤮
-
 HEART_EMOJI = ["❤️", "♥️", "💙", "💜", "💛", "🧡", "💚", "🩵"]
 
 
@@ -36,7 +34,6 @@ def meme_reaction_keyboard(
 ):
     like = random.choice(HEART_EMOJI)
     dislike = "⏬"
-    # like, dislike = "👍", "👎"
 
     def reaction_callback(reaction_id: int) -> str:
         callback_data = MEME_BUTTON_CALLBACK_DATA_PATTERN.format(