-There are really only two times you'd skip a lynch. Day 1 and if you can reasonably assume that you are in mylo (and to just cover bases since this is our strategy section, MYLO stands for Mislynch and Lose meaning you must lynch scum or the game is over BUT no lynch is a valid option that will not end the game as opposed to LYLO which is Lynch or Lose meaning you must lynch scum today or the game is over). So the whole day 1 topic is kind of a crap shoot which is why it gets brought up so much. To be as unbiased as possible, the more vanilla a game is, the more you should err to lynching on day 1 while the more role madness a game is, the more you should err to no lynch on day 1. The reason for this is that in a vanilla game, the ONLY information you will be getting is vote and lynch records so skipping day 1 means that you've just skipped a whole day where you can analyze votes. Meanwhile on the role madness side of things, the consideration is twofold, 1) getting scum day 1 is hard, that usually involves a lot of luck or a colossal fuck up from someone, between all scum not really ever needing to bus day 1 and just the pure odds, you are just more likely yo lynch town and the closer you are to role madness that also means you are likely to lynch a town PR 2) you will likely have access to other information sources in these games, a movement detecting role or investigator of some sort, they will provide you info in addition to normal voting records and flips meaning you can get away with skipping day 1, you are also providing them with extra cover since that's one more unlynched body that scum may choose to kill instead of the cop or tracker or whatever over night (this is negligible though because you are also leaving 1 extra body to be tracked or investigated or whatever and while town checks are useful, scum checks are really the main prize)
As far as mylo/lylo, that's pretty self explanatory, even being unbiased, I would say you just vote no lynch in mylo 9 times out of 10. The only time it's the wrong call is if scum has an extra vote or extra kill that they have been saving all game until this final moment and that is such a rare case that you should never be planning for it. You can also fuse these two ideas together though it doesn't work much in our community since our games tend to have multiple killing roles in them. If you enter a game that has an even number of players and you can reasonably assume there will be no extra kills (or an even number of extra kills, maybe you are a two-shot vig for example) then you can kill two birds with one stone, voting no lynch day 1 here will cause the player count to go odd instead meaning you would always hit lylo instead of mylo.. Again, that's probably useless info for our community because of how unpredictable our kill PRs are but it's something you can always keep in mind.
-WIFOM isn't going to get much tbh. There is no easy answer here. Everything is always worth discussing because it's a game of deduction but I promise you that whatever obvious conclusion you have reached, someone else has reached the totally opposite conclusion and is just as sure in their conclusion as you are in yours. Add in scum trying to purposely fudge things up and really, that's just the game of mafia.
-Policy lynching is a hard one. The easy answer is that it depends on the skill level of your community. If the player's are generally good all around then policy lynching is the right call, it's likely you are seeing bad play due to being scum/nerves/having to protect a teammate/etc. If the players are generally bad in the community then policy lynching tends to be a worse call. You are more likely to hit town in this case and if you do even find scum, it was probably more due to dumb luck than actual caught out play on their part. There's more nuance of course, communities don't tend to have a blanket skill level and you'll find a bunch of people fall all along a curve. There's probably also more to be said about what if your community is particularly good at the town deduction side of the game but bad at the acting scum side or vice verse. Honestly, the best advice is probably to just say if you find bad play or something worth policy lynching, maybe use that as a jumping off point to focus on that player. Not to tunnel them per se but use that as an indicator to go read the entirety of their post history. See if you spot more obvious scummy behavior like inconsistencies in reads vs. votes, protecting scum that may have been revealed already in game, etc. And if you see something (or if the issue is they don't post enough), vote for them, see what shakes loose but keep an open mind at how they react.
That said this is also probably worth a little blurb, "bad play" worth policy lynching: Inactivity, whether that be overall low post count or more focused inactivity (here a lot during the first 24 hours but gone for the next 48 or similar things) orglaring anti-town stuff (voting for them self, claiming a PR but never using or forgetting to use said PR, frustration leading to exiting the game, literally or figuratively). These things are not scum tells, they CAN be but that's where the extra re-read will help because you should be voting for JUST these things unless it's really early days and there's nothing else happening for whatever reason. "Bad play" that can actually be scum tells: defending scum, pushing a lynch onto town, scum reading someone but never really committing to the vote on them until they are forced to (and bonus if the person than flips as scum that they wouldnt commit to), tunneling. Now, all that said, there's obviously nuance here, town will be wrong. That's just the facts of the game, the person right about literally everything probably has to be scum just on the basis that scum has so much extra info. You're looking for consistency here to really read it as a scum tell. Did they refuse to completely commit to a vote that turned up scum and now you find them being non-committal to their reads again? That might be worth a vote. Did someone spend a big part of the game tunneling and after being wrong jump right into another tunnel without evaluating the game state? The pattern of behavior is probably the simplest way to answer this whole question but I hope the rest of this section wasn't a waste of time to read.