The previous article covered some books intended for more general use, including style guides and student grammars. This covers the reference grammars which are available.

Of largely historical interest

Henry Sweet (1892, 1898) A New English Grammar, Logical and Historical, Clarendon Press. Part 1 (Introduction, Phonology, Accidence) and Part 2 (Syntax) are available at archive.org: vol 1 here, and vol 2 here. Although there are frequent references to Old English, Sweet is remarkably non-prescriptive, and still makes fascinating reading.

Otto Jespersen et al. (1909-1949) A Modern English Grammar on Historical Principles, reprinted from 1954 edition by Routledge. Volumes 1 (Sounds and Spelling), 2-5 (Syntax), 6 (Morphology), 7 (Syntax). It is also worth adding his separate work on Negation, perhaps. Still a wonderful work to refer to, with remarkably detailed historical information. However its current reprinted version is an expensive source, and it is not yet out of copyright so not available (legally) electronically.

H Poutsma (1914-1916, 1928-1929) A Grammar of Late Modern English, P Noordhof. 1st and 2nd edns. Parts 1 (The Sentence), and 2 (The Parts of Speech). Builds on Sweet and Jespersen, and describes the language in the 1920s. Available at archive.org: part 1a here, part 1b here, part 2a here, part 2b here, and part 2c here. Aimed mainly at Dutch students, but a very extensive work of great interest.

E Kruisinga (1931-2) A Handbook of Present-Day English, 5th edn, P Noordhof. Part 2 (Accidence and Syntax), of which there are 3 vols. Available at archive.org: part 2a here, part 2b here, and part 2c here. Another monumental work.

Randolph Quirk, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (1972) A Grammar of Contemporary English, Longman. ISBN 0 582 52444 X.
The benchmark for many years, this was the first truly modern grammar, and its usage information was based on the Survey of English Usage conducted by University College London (UCL). Running to over 1100 pages, it was in a class of its own and still makes interesting reading, as it reflects British and American English usage in the 1960s. My copy, bought in 1976, cost the princely sum of £14.

A Big Four

Randolph Quirk, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (1985) A Comprehensive Grammar of the English Language, Longman. ISBN 0 582 51734 6. Total 1779 pages. This is effectively the second, rewritten version of their 1972 grammar (above).
There are appendixes covering word formation, stress, rhythm and intonation, and punctation. There is an excellent and extensive bibliography, but no annotations to guide any quest for more information. Its single index is very thorough but infuriatingly refers to section numbers rather than pages.
Although another big step forward from their original grammar, this has dated considerably and is probably best reserved for use on older matters.

Douglas Biber, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan (1999) Longman Grammar of Spoken and Written English, Longman. ISBN 0 582 23725 4. Total 1204 pages. This is the parent of Biber, Conrad and Leech’s Longman Student Grammar of Spoken and Written English.
Based on the Longman Spoken and Written English (LSWE) corpus, this is probably the best corpus-based grammar at present. Coverage includes American and British English, spoken and written, in a range of registers, including published fiction, news, and academic prose. Throughout the text, there are detailed presentations of corpus findings, with bar tabulations and formal frequency charts. These compare alternative constructions across different registers and between American and British usage, making this a uniquely detailed volume.
The fifth section, containing four chapters over more than 200 pages, focuses on word order and syntactic choices such as existential there, the grammatical marking of stance across different registers, lexical expressions such as idiomatic phrases, and the grammar of speech, including dysfluency and error. Although these topics are selected and not perhaps as comprehensive as those in previous sections, they add particular value to this grammar.
Indeed that is perhaps the biggest criticism of this otherwise excellent grammar: that it is very strong on usage which is sufficiently frequent to feature in its corpus, but lacks detail on less frequently-encountered topics. You would therefore be ill-advised to make this your sole reference grammar.
The content is well laid-out, although some of the charts are a little small even for the myopic reader. Although no colour is used in the text, tables are well presented and clear, and sensible use is made of layout and typographic devices.
There is an appendix covering contractions, endnotes largely concerning the corpus source, and an extensive bibliography which classifies entries but gives no further notes or information about them. There are two indexes, lexical and conceptual, which are comprehensive and excellent in use, particularly as they refer to page numbers.

Rodney Huddleston and Geoffrey K Pullum (2002) The Cambridge Grammar of the English Language, Cambridge UP. ISBN 978 0 521 43146 0. Total 1842 pages, and now most surprisingly available for free download from here.
It is no relation to the other Cambridge grammar, and given the differences between them, you would have thought that more distinctive titles would have been desirable.
It is hard to know where to start to describe this grammar, other than by nominating it as the closest that anyone has come to perfection. Its two principal authors enlisted the involvement of a team of experts to ensure that individual topics such as morphology were covered fully, but retained overall control of the work, which has ensured consistency of approach and clarity throughout.
This is even more comprehensive than Quirk, Greenbaum, Leech and Svartvik, and I have yet to find any point of English grammar on which it is not detailed. For instance, of the ‘syntactic nuts’ that I have been describing here, this is the only grammar which often gives more detail than my original sources. It is the only grammar in which you will not just find one mention of resultatives, but five! The only topic on which I have so far found it less than definitive is the order of geographicals in place names, which is perhaps on the margin of grammar in any case.
Layout is excellent, although there are relatively few figures (this not being based on any specific corpus), mainly the occasional structural diagram. Colour is used to distinguish headings, and there are some boxouts which appear in pale blue; the latter remain very clear, and even in poor light I have never struggled to read its pages. Examples are not attributed to source, though. There are occasional footnotes, which appear on the same page for ease of access, and are not relegated to the back of the book.
There is an excellent guide to further reading, prefacing an extensive list of references. There are two indexes, lexical and conceptual, which are comprehensive and a joy to use, particularly as they refer to page numbers.
This is the established standard, whether you are a linguist or user of the language. It is not only the first reference which I go to now, but is also one that I find myself quickly becoming engrossed in. I am sure that reference grammars are not supposed to be hard to put down, but this one is.
A list of errata is available here.

Ronald Carter and Michael McCarthy (2006) Cambridge Grammar of English, Cambridge UP. ISBN 0 521 85767 8. Total 973 pages, plus a CD that only runs on ancient versions of Windows.
Aimed at the non-linguist advanced learner, and with a minimal linguistic overhead which should make it more accessible to those trying to improve their English rather than learn linguistics. It adopts unconventional stylistic conventions such as striking out ‘incorrect’ examples rather than using an asterisk, which would be fine for those not used to other reference grammars. There is good use of colour for cross-referencing, emphasis, and tables, without impairing legibility, and contents are generally laid out well to make their access easier.
Based on the British English conversational CANCODE corpus, its emphasis is on current usage in spoken English, although there is also adequate coverage of written English. Examples are not referenced to sources, and in general details such as footnotes and citations are avoided, which keeps the text clean and clear. However this also makes it impossible to follow up issues that might arise. The account is well-structured for the advanced learner, although it lacks finer details which are infrequently encountered in spoken or everyday written English.
There are appendixes covering word clusters, punctuation, spelling, numbers, time, measurement, nationalities, irregular verbs, and American English. A long glossary is aimed at non-linguists. A brief bibliography is confined to papers on the CANCODE corpus, and does not suggest further information about English grammar, for example. Its single index is thorough but infuriatingly refers to section numbers rather than pages.
This is the definitive reference grammar for the advanced learner, whether English is their first language or not. However in terms of further development – linguistic, historical, or more detailed – it is a bit of a dead end.

Which should I buy?

For anyone not wanting to mess with linguistics, and for whom accessibility is most important, Carter and McCarthy is an excellent choice if you are happy with predominantly spoken and everyday British English. However you then sacrifice any ability to take topics further, without resorting to another grammar.

The original and extended versions of Quirk, Greenbaum, Leech and Svartvik still have some appeal and considerable value, and the later edition is still eminently usable, but they have been superseded now.

Biber, Johansson, Leech, Conrad and Finegan is an excellent corpus-based grammar which should appeal more to those who like usage data combined with linguistics. However being corpus-driven it rapidly thins out when covering less common constructions, and you would then need to refer to a more conventional reference grammar, either Quirk, Greenbaum, Leech and Svartvik, or better stiil, Huddleston and Pullum.

Overall, then, if you want a grammar that is truly comprehensive and insightful, containing linguistic detail, Huddleston and Pullum still stands head and shoulders above the rest. If you want to add corpus data and further detail on spoken English, then Biber, Johansson, Leech, Conrad and Finegan would be an excellent supplement.