This is quite simply a side-effect of the basic system that Pathfinder 1e (and D&D 3.5e before it) is built on.

There are a couple of tricks that can be used to deal with this...

Give the players more points to point-buy with.

Sounds stupid, but this (usually) is a reasonable solution. As others have pointed out, there’s a soft requirement for a CON of 14 in PF 1e to just be able to survive (barring having a very high AC or being otherwise very well prepared). The ‘normal’ point-buy option makes meeting that and other class requirements challenging, especially for certain classes, so just bump things up to the high-fantasy or epic-fantasy point-buy levels and nudge your players at character creation to ensure they have good contingencies for dealing with serious damage, be it very high AC (20 or higher at level 1 is usually enough to make things a non-issue), a good CON, or some other contingency to allow them to take a few solid hits without outright dying.

Skip the first few levels.

This is actually something I suggest for almost any D20 derived system. Level 1 is horrendously binary, being either mind-numbingly boring or downright deadly with essentially no in-between. So why not just skip it? The rules in pretty much any system cover starting above level 1 (it largely amounts to creating a level 1 character stat-wise, leveling it up to the appropriate level, and then giving the player a set amount of gold to work with for character creation). You can come up with all kinds of story reasons for starting above level one. For example, I’m putting together a campaign where the party are mercenaries hired to do the main quest by a major international alliance, so it makes no sense there for them to be starting at level 1.

This will quite often make most veteran players happier too because it lets them get right into the fun parts of the game instead of having to hope a bad roll means they need to build a new character.

Have everybody keep backup characters in-reserve.

This is, honestly, the single best approach I’ve found to this issue. Have everybody create a secondary character and have it gain XP concurrently with their main character. When their main character dies and they either chose not to be revived or can’t be revived, they kit out the secondary character with equipment based on expected wealth for it’s current level, and that character then joins up with the party next session.

This reduces the overhead of creating a new character and if sold right helps soften the blow of losing a character the player cared about (especially if you can work with the players to come up with story hooks for why the backup might join the party). VTTs like Roll20 even make this easy, because you can keep the backup sheets hidden from everybody but the GM and the player who the character is for until they’re needed.

Note that usually when I do this, I insist that the backup not be a clone of the existing character. I have no issue with them filling the same role, but they should not be a cookie-cutter copy of the existing character. This helps ensure that character death is still meaningful, which helps keep the party playing sensibly instead of treating everything like a suicide mission and not caring about death.

Just fudge the rolls.

You’re the GM. The dice don’t decide how things go, you do. So what if a monster rolled a crit and then went and rolled max damage on it? You’re the GM, you can just narrate it to sound spectacular and then have it do almost nothing (and then the character gets to quote Monty Python, because ‘It’s only a flesh wound!’). The goal is to have fun, and most people playing TTRPGs aren’t looking for a no-holds-barred roguelike experience.

As far as players ‘expecting’ it, that should not ever be a problem. They should not be expecting you to fudge rolls, otherwise they’re either way too distrusting/suspicious or you’re doing it way too often.

Aside from this though, there’s a very strong argument for not showing players any rolls made by anybody other than themselves. Ignoring the fact that Roll20 always shows the full breakdown of the roll in the tooltip (which gives the players way more info than they normally would have in a regular game, I have had issues before with people metagaming based on this breakdown), the simple fact of seeing the value on the die and knowing whether it succeeded or failed is enough for anybody who’s paying attention to figure out the associated total modifier after only a couple of rolls (good players can get a reasonable estimate after only 1-2 successes and 1-2 failures), and that level of detail isn’t exactly something the characters would have access to, which leads to the complicated issue of potential metagaming.