I have a RB2011UAS RM and when any user after total of 112 are connected they show in red and invalid under /ip addresses and cannot connect to internet, if I remove a user and then remove the one in red, the one in red connects OK, if I then connect the one that was OK invalid.
What IP pool are you using? Are there any more IPs available in the pool?
There are 10 IP’s left in the pool 3 are reserved, almost all the IP’s are statically assigned via the Radius server Framed-IP Address attribute, these were all assigned top down 254-14, pool runs from .4 to .13, inclusive therefore the pool cannot overrun the static IP’s
Because the MikroTik takes IP’s from the pool top down, as one is assigned I make it a static with the Framed-IP-Address attribute and alter the pool accordingly.
I have assigned 112 users statically when the 113 attempts to connect, it connects but becomes invalid, it is invalid either statically assigned or dynamically assigned, it makes no difference
Activate the log for pppoe in /system log and check what the log says about it.
Everything looks OK in the Logs, user 112 and 113 are identical except for user name and assigned IP, what is different is the router table, no route has been assigned for user 113.
I can switch off user 112 and reboot user 113 so it becomes 112, and it works, switch on the old 112, it becomes 113 and invalid and no route.
At the moment it is fully populated with 112 users so I cannot emulate user 112 only add router to emulate 113, like I could earlier when not filly populated.
I think there got to be something on the log, it can not be all ok. Weather it is a problem when creating the tunnel or when trying to get the IP. If the IP are handled by radius, try once to assign the IP from the router itself.
Already tried creating a secret in the PPP section and removing it from the radius, I tried it with both Dynamic assigned IP from pool and static assigned IP address, exactly the same problem under addresses it shows DI and it is in red and no route created, so it nothing to do with radius, again everything looks the same in logs for the last 2 users, it looks as if there is a limit to the number of dynamic routes that can be created.
In the middle of the night I backed up all the configuration of the RB2011UASRM by exporting each section to .rsc files and restored on to my old RB1100AHx2 with 5.25 on it ( if it is not broke do not fix ) having first made sure the interface names matched those of the RB2011 done this a few times over the years.
At 4:00AM I swapped the RB2011 for the RB1100AHx2 and low and behold no problem with the 112 user limit.
The question now is this, is it a problem specific to the RB2011 or is it a problem with later firmware, I have no intention of spending hundreds of £’s on a core router until I get an answer.
Nice you found a way out of it 
The 2011 has less RAM and CPU then 1100AHx2, so it could be hardware related. This can be easily verified by looking at system resources on the 2011 (although I hardly believe you have reached the ceiling of the resources of it). Otherwise it could be software related.
The CPU was at 27% that is all not over worked just because they are connected does not mean data going through.
There is a limit of PPPoE connections on level 5 license, it is supposed to be 500 on Level 3 200 not 112, It is working fine on the RB1100AHx2 with the same configuration, so it is not a miss-configuration.
There obviously a serious bug somewhere in the system, I do not know if it is hardware or firmware.
My RB1100AHx2 is old and I want to replace it with a core router but I am very reluctant now to risk it, if it is firmware related it will be the same on the core routers.
The RB2011 was temporary because the RB1100 was still on the old link when we were changing over to our new suppliers fibre cable, it was intending to purchase a core router and keep the RB1100 for the backup link, I not sure if I want to risk it.