Has the MSO/LA improved any? That's a big feature to me and I have every intention of upgrading to it if it is not crippled.
Are there any big show-stopper bugs still out there?
The integration of the MSO option with the SDS1004X-E DSO is not great, so you can call it crippled indeed, when compared to the fully integrated solutions in SDS1000X+ and SDS2000X.
Last time I've checked it about 9 months ago, there has been some 7ns skew between analog and digital channels and the digital deskew parameter was not preserved upon a power cycle. I expect this bug to be fixed by now though.
Mixed channel Pattern Trigger is not supported, so it has to be either an analog or digital pattern.
Furthermore, both History and Zoom don't work when digital channels are enabled. Since most people want the MSO for decoding serial buses nowadays, this means you cannot decode long messages, because the decoder line at the bottom of the screen will become unreadable due to the lack of space. Using analog channels, you can easily deal with that by just entering zoom mode and take a closer look at the part of the message you're interested in. As it is now, when using digital channels you're almost limited to the capabilities of a Rigol 1000Z which only decodes the screen buffer anyway
The limitation described above only applies in run mode though. You can still capture a long message and then zoom in and navigate through the message while in stop mode. Also the list view works during Run even for very long messages, but that's rather pointless as you cannot closely examine the corresponding waveform – you might just as well use an LA instead.
Other than that, it's not all that bad. Here is an example how a 100MHz SDS1104X-E is capable of capturing a 320MHz sinewave with only 1.5Vpp amplitude on a logic channel of the SLA1016:

SLA1016_D12_Sine_320MHz_1.5Vpp
A 4ns wide pulse – heavily distorted when viewed on the analog channel at just 100MHz bandwidth, but nicely captured on the digital channel:

SLA1016_AD_Pulse_5V_4ns_probe
16 digital channels, grouped into 2 parallel buses, captured at 1GSa/s with 14Mpts record length. This also demonstrates how the decoded values become unreadable in the busy regions, where we get just blue bars. In stop mode, zooming into these regions wouldn't be a problem though. I just don't happen to have a screenshot at hand for that.

SLA1016_Bus_Select
Finally a digital pattern trigger. Two buses again, the first (lower) one decoded binary, the 2nd (upper) one decimal:

SLA1016_Trig_Pattern_1us