First of all, to understand my level of experience: I've been mapping in Hammer since 1998 and released my first few "successful" maps in 2001, although what "successful" means changed a lot over that time. Back then, i was quite happy with about 10000-15000 downloads and a release on a CD in a gaming magazine with 1/3 of a page within an article back when download speeds where so slow that custom content in gaming magazines was still a regular kind of thing.
I also wrote map reviews for planethalflife's Biohazard when map review sites still existed, so i saw lots and lots of custom content back then.
I got out of the loop shortly after CS:S was released and only got back in late 2014. There are still a few things i learned back then that i can apply now, but a lot of new things were added. I know how to use func_detail from using func_wall back then, even though the reason for adding them is quite different. Back then it mostly was because a brush touching another cut one of them down into more polies, and there was a pretty hard low poly limit for software rendering. Also, the overall maximum level of detail became way more relaxed with Source, but the stylized level of detail for TF2 fits the old HLDM style quite well, so i can still apply knowledge there.
Currently, when i'm creating a map, i try to learn as much as somehow possible from it. That's the reason why i put a lot of work into koth/arena_temper although i knew the map failed on a fundamental level for quite a few reasons (the layout of the map didn't really even line up well with 2003 standards.) I also have to line up my mapping hobby with a demanding full-time job (although it allows me to do some mapping during downtimes.) Because of these reasons, i'm more of a "few large projects" kind of guy - But that doesn't mean i haven't or won't scrap a project if i think there is not much left to learn from it if i think my new knowledge is best applied to a new project because of fundamental flaws.
Considering feedback - The public testing that happens on TF2M wasn't really a thing back when i started mapping. Gamedays, which happen consistently every week here were a big thing that happened maybe bimonthly. Because of that, there were way less versions released and maps changed quite a lot more between versions, sometimes so much that it was impossible to fall back to the old version should something not work out the way you intended it too, so the level of failure was much bigger than it is now, with much more time put into the map before you realized that it did not work.
The feedback you get from imp and gamedays is a huge privilege that you should use as much as possible. I know that it sometimes can be hard listening to the flaws of your map over and over, i had my own share of problems with that. People will have different opinions about what works on your map and what doesn't, and it's important that you have your own vision for the map, because if you try to listen to everyone, you will create a map that makes everyone equally unhappy. Learning how to interpret feedback and results from the tests and implement changes according to them is as much as a learning experience and as important as learning basic layout, detailing etc - And it's also the field that is the newest for me and easily my weakest spot in mapping.
But as i learned from the release of Hierarch, even a map that received "mixed" (in steam terms) feedback from the testing group can be kinda successful upon release. Although i know that the Steam Workshop is a fickle beast and getting top 3 is basically just a self-fulfilling prophecy once you get there, i consider 1500+ subscribers, 98% positive rating over 500 ratings and 10 days within the top 3 most popular maps a success for my first full-fledged release, and that with the testing feedback being quite negative over the last 5-6 tests. So, take that however you want - i know i will.