The dumpster fire that was Comcast internet (100/20) at the office I work with was eventually replaced with Comcast's fiber service (100/100) at several times the price. Expensive, but it does have an actual SLA, and has proven completely reliable so far. The previous service was plagued by:
First, absolute garbage ISP-provided routers, required due to their asinine scheme for implementing static IPs:
Cisco DPC3939, 3941: Defective Intel Puma 6 chipset, and everything that goes along with it. WiFi was, however, pretty decent. Drops connection randomly; web pages randomly take 30+ seconds to load due to retries. RDP and other real-time protocols are basically unusable over this turd. Seriously, f*ck this router. A large part of its issues seemed to stem from a bad reaction to erratic network connectivity on Comcast's side. When the pipe was calm, as at night, things weren't too bad.
SMC D3G: Broken IPv6 support (only one IPv6 host will work properly), only 4 upsteam/4 downstream channels. Could not have been used for service beyond ~150mps as a result. Had to disable IPv6 on ~40 computers to keep Outlook from having a fit (it was the only application that seemed to be unable to cope with it). This one was mostly okay, but was not fast. Eventually died outright, leading to...
Netgear CG3000: Weird issues that felt similar to the Ciscos, despite using a Broadcom chipset, but far less severe. Forgets custom DNS servers for DHCP clients, and starts handing out Comcast's about 1 minute after booting, regardless of what is shown in the UI. First one we got (used) died in a month or so. Second one had all of the same basic issues, and they refused to provide a brand new one. Locked up randomly, sometimes 2 or 3 times in a day, and sometimes went weeks without locking up. Drops packets reliably, once per minute including to the LAN IP, followed by a few packets with 200-2000ms responses. Performance degrades over the first hour or two of use, and then stays bad. Only router I've ever had to use that has consistently erratic ping times on its LAN interface, as well as packet loss to the LAN interface.
Second, we were on a seriously overloaded network segment which would show latency and packet loss that made for some very jagged and colorful Smokeping graphs with any of the three router types, especially the Ciscos. While available bandwidth would typically stay above ~70%, packet loss of 1-2% was frequently observed, and would often spike to 5-10% for minutes at a time. This was almost always coupled with latency spikes to 100-5000ms. This seemed to occur pretty much daily, with all of the routers, though the SMC had the least issues (rarely more than 0.5% loss, and rarely more than 200ms). Note that these values are for pings to several of Comcast's core IPs such as their DNS servers and the nearest-hop router.
Finally, Comcast support could do little more than "send a signal" (i.e. remotely reset) the router or schedule a tech to come out. Multiple technicians admitted openly that the particular plant was seriously overloaded, and submitted trouble tickets, but it was never fixed in the 3 years we had it. Even showed me some of their network diagnostic software views. Signal strength to/from the router was perfectly fine. Their network and their corporate constructs are, to use an expression I've become fond of, a dumpster fire, clown shoes, sh*tshow.
Business Class, my ass.
Biting the hand that feeds IT
