diff --git a/NEWS.md b/NEWS.md
index f908539050..c925bac7ee 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -59,6 +59,8 @@
 
 7. `as.data.table()` failed with `.subset2(x, i, exact = exact): attempt to select less than one element in get1index` when passed an object inheriting from `data.table` with a different `[[` method, such as the class `dfidx` from the `dfidx` package, [#4526](https://github.com/Rdatatable/data.table/issues/4526). Thanks @RicoDiel for the report, and Michael Chirico for the PR.
 
+8. `rbind()` and `rbindlist()` of length-0 ordered factors failed with `Internal error: savetl_init checks failed`, [#4795](https://github.com/Rdatatable/data.table/issues/4795) [#4823](https://github.com/Rdatatable/data.table/issues/4823). Thanks to @shrektan and @dbart79 for reporting, and @shrektan for fixing.
+
 ## NOTES
 
 1. New feature 29 in v1.12.4 (Oct 2019) introduced zero-copy coercion. Our thinking is that requiring you to get the type right in the case of `0` (type double) vs `0L` (type integer) is too inconvenient for you the user. So such coercions happen in `data.table` automatically without warning. Thanks to zero-copy coercion there is no speed penalty, even when calling `set()` many times in a loop, so there's no speed penalty to warn you about either. However, we believe that assigning a character value such as `"2"` into an integer column is more likely to be a user mistake that you would like to be warned about. The type difference (character vs integer) may be the only clue that you have selected the wrong column, or typed the wrong variable to be assigned to that column. For this reason we view character to numeric-like coercion differently and will warn about it. If it is correct, then the warning is intended to nudge you to wrap the RHS with `as.<type>()` so that it is clear to readers of your code that a coercion from character to that type is intended. For example :
diff --git a/inst/tests/tests.Rraw b/inst/tests/tests.Rraw
index b230f5f2e8..b7a444864c 100644
--- a/inst/tests/tests.Rraw
+++ b/inst/tests/tests.Rraw
@@ -17373,3 +17373,7 @@ class(x) = c('foo', 'data.frame')
   as.list(x)[[i]]
 }
 test(2173, as.data.table(x), data.table(a=1:5, b.V1=6:10, b.V2=11:15))
+
+# rbind two length-0 ordered factors, #4795
+DT = data.table(A = ordered(character()))
+test(2174, rbind(DT, DT), DT)
diff --git a/src/rbindlist.c b/src/rbindlist.c
index bb42502be6..5dab7fff51 100644
--- a/src/rbindlist.c
+++ b/src/rbindlist.c
@@ -273,7 +273,7 @@ SEXP rbindlist(SEXP l, SEXP usenamesArg, SEXP fillArg, SEXP idcolArg)
   for(int j=0; j<ncol; ++j) {
     int maxType=LGLSXP;  // initialize with LGLSXP for test 2002.3 which has col x NULL in both lists to be filled with NA for #1871
     bool factor=false, orderedFactor=false;     // ordered factor is class c("ordered","factor"). isFactor() is true when isOrdered() is true.
-    int longestLen=0, longestW=-1, longestI=-1; // just for ordered factor
+    int longestLen=-1, longestW=-1, longestI=-1; // just for ordered factor; longestLen must be initialized as -1 so that rbind zero-length ordered factor could work #4795
     SEXP longestLevels=R_NilValue;              // just for ordered factor
     bool int64=false;
     const char *foundName=NULL;